But what about seeing something when you think you are totally blind? What about navigating around obstacles that you cannot see and aren’t even expecting?

The article is talking about blindsight, a condition where people who have sustained brain damage to the visual areas of the brain report being blind, but can still respond to some visual stimulus. For example, they can tell whether an image is being presented or not, whether it’s on the left or the right, or if it’s horizontal versus vertical stripes—all while reporting they don’t see anything. In one dramatic demonstration, a blindsighted man walking down a hallway navigated around objects in his path.

Because blindsight is purported to be a dissociation between conscious awareness and visual ability, it’s captured imaginations, being featured in many philosophical debates. It even inspired a science fiction novel, where author Peter Watts was inspired to create aliens who are not conscious but still act intelligently. Fittingly enough, the novel is titled Blindsight.

I’ve long been aware of blindsight, and generally accepted the story that it’s a case of dramatic dissociation between awareness and sight. I’ve pointed to it often as an example of “one of those weird neuroscience cases that tell us something about consciousness”.

However, I had never read the literature on blindsight until recently, and it was a bit of a surprise that I came away skeptical that it’s anywhere near as special as it’s sometimes reported to be.

Subscribe now

What is blindsight?

Blindness comes in different forms. Most types involve some issue with the eye itself: Cataracts can cloud the eye’s lens, the blood vessels in the retina can be damaged, or the optic nerve that carries signals from the retina to the brain can be damaged.

Cortical blindness is different, it involves damage to the part of the brain that processes the visual information from the eye. The eye itself is fine, it’s the brain receiving the signal that isn’t working.

Two bombshell papers in 1973 and 1974 described experiments done with cortically blind patients who were able to do simple visual discriminations in the lab, despite reporting not being able to see anything. This was followed by decades of research on blindsight patients and intense debate about it and its implications.

The idea of us being able to take an action without consciousness is nothing new—reflexes exist, after all. What’s striking about blindsight is it involves discriminations that are normally conscious and can be acted on flexibly.

The initial reports of blindsight were met with skepticism, but it’s generally accepted now and appears in textbooks. Still, there’s significant controversy around it. A 2021 paper argued that blindsight is actually just “qualitatively degraded conscious vision”. This sounded ridiculous to me when I first came across the paper, but I now at least lean in the direction that it’s got things right.

Subscribe now

Discrimination Without Awareness

It might seem crazy to say that blindsight is just degraded conscious vision. The patients report not seeing! Their vision is clearly not conscious!

To get a bit of a handle on why this isn’t that unreasonable, we need to go into the world of psychophysics, which is my favorite name for a subfield because it sounds like it’s straight out of a science fiction novel. Psychophysics is the study of the relationship between physical stimuli and their perceptions. A canonical type of experiment is finding the threshold at which a stimulus can be detected. For example, how little light can you use before someone no longer detects it? Or how much of a brightness difference does there need to be between two lights for someone to notice the difference?

One common kind of stimulus used in psychophysics (and vision research more broadly) is the Gabor patch: a patch of oriented, repeated bars that fade out around the edges. Psychophysicists can alter the properties of the patch to see how it impacts how well we can detect it. For example, you can lower the contrast—see the example below, where on the right is a high contrast Gabor patch and on the left is a very low contrast one.

You can play around with the properties of Gabor patches yourself here. I generated the left one by testing how low I could put the contrast to feel like I’m just at the edge of what I detect. On my screen, in my particular lighting, I can somewhat reliably see that the left isn’t just a grey circle, but other times I feel like maybe it’s my eyes playing tricks on me.

Blindsighted individuals aren’t making discriminations that are that subtle. If the discriminations they make in lab tasks were conscious, they would still be well below the threshold of legal blindness. Here’s the sort of discrimination they would struggle with: high contrast bars that are either horizontal or vertical. When asked “is it horizontal or vertical?” they might get it right 60-70% of the time.

Now to get a sense for how psychophysics experiments work, imagine being shown, possibly around 100 times, different Gabor patches around the threshold of what you can see, and being asked their orientation. Then, separately, you rate whether you saw it or not. It’s not hard for me to imagine that, on many trials around my threshold, I might have the “eyes playing tricks on me” feeling, and respond that I didn’t see it.

In “signal processing theory” terms, what I’m describing is called a conservative detection criterion—my ability to make discriminations outstrips my willingness to report my awareness. Research indicates it’s common for people to have a conservative detection criterion when they’re at the edge of their ability to discriminate. That is basically the deflationary case against blindsight—that it’s basically the same experience as anyone who is dealing with a stimulus right at the threshold of what they can detect.

You can see the association between their “detection sensitivity” (how well someone is discriminating things) and their “detection criterion” (how willing they are to report awareness) in the figure below. As detection sensitivity goes up (as we move to the right), detection criterion tends to go down. When a task is easier, we’re more willing to say we’re aware.

The red point is a blindsight subject, the other points are normal healthy people performing a difficult discrimination task.

The point is that the red dot here doesn’t stand out that much. It’s slightly above the others, indicating a slightly higher conservativeness than you would expect, but it’s not a crazy outlier.

Simple factors could explain why blindsight folks are more conservative—for example, since many of them have vision in parts of their visual field (or previously had vision before the injury that led to their condition), they might be comparing “seeing” to that vivid visual experience and acknowledging they don’t have it.

This is the kind of grey area blindsight studies are sitting in. It isn’t a patient stating clearly “I see nothing” but then being able to detect the stimulus perfectly 100% of the time. It’s being at the edge of what they can detect, and then showing they detect something fairly well despite reporting not being able to see it. That’s when subtle things like how you ask a question start to matter.

Subscribe now

It Matters How You Ask The Question

Is there a difference between asking someone “Did you see something?” versus “Were you aware of something?” In everyday life, those seem pretty similar—usually we’re dealing with clear, vivid sights, so seeing something implies you are aware of it and vice-versa. But it turns out subtle differences like this can elicit very different effects.

The below figure was generated during a test of peripheral vision in a blindsighted person. The white circles indicate where the person gave a positive response, the black (you can’t really tell but they’re also circles) when they gave a negative response. Only the right visual field is tested, which is why there are no circles on the left.

The difference between a and b is simply phrasing: a is “Did you see something?” and b is “Were you aware of something?”

Clearly, the threshold for when this person says they are “seeing” something is higher than when they say they are “aware” of it. If what they’re aware of is vague, they might be willing to say they are aware of something but don’t see it.

In a typical blindsight study, participants are given two possible responses, was the stimulus seen or unseen. But a couple of studies have now tried asking on a more graded scale, the 4-point Perceptual Awareness Scale: no experience, weak experience, almost clear experience, and clear experience. In one study involving a blindsighted woman, using the traditional “yes-no” question evoked the usual blindsight result (above-chance detection when reporting “no” for seeing). But when they switched to the 4-point scale, she was at chance for “no experience” and only above chance for “weak experience”.

In other words, it seems she was lumping “weak experience” in with “no awareness” when she was only given two options. Which makes sense—there’s a big difference between very obvious awareness and the sort of “my eyes might be playing tricks on me” awareness that sometimes happens at the threshold of what we can detect. It seems “awareness” comes in degrees.

There’s long been a distinction between “Type I” and “Type II” blindsight, where Type I is the kind you typically see mentioned (detection with no awareness) and Type II involving detection with some awareness. This already makes blindsight a bit more of a gradient than is often acknowledged, but interestingly, with training, some people both increase their ability to detect and their awareness, going from Type I to Type II, and some eventual full visual experiences.

I don’t want to give the impression that these simple measures mean the issue is clear-cut and, actually, if you change the way you ask a question, it completely undermines the effect. Blindsight patients are rare, and while there has been a lot of research on them, they tend to be on very few subjects (often one), and methodologies are varied. But the important point is that in the popular imagination, blindsighted people are described as having absolutely no visual experience, whereas the reality seems to be that they often have some, and the line between seeing and not seeing isn’t as clear as you might think.

Subscribe now

What is it like to have blindsight?

The picture I get from reading about the above research is that blindsight is much less like a reflex, where there’s no awareness, and much more like seeing in a dense fog. But that doesn’t quite get at it, because the kind of awareness blindsighted folks have is different from typical vision.

First, as mentioned above, the kind of discriminations they’re making are not good. The typical blindsight result is that they can tell apart big, obvious stimuli at somewhat above chance levels.

It’s been suggested that while some low-level features can be detected by blindsighted people, they can’t combine those elements into forms and objects. So it isn’t like seeing shapes in a fog, but more like being able to dimly sense the different components of a shape but not seeing how they come together.

It’s typically reported that blindsighted folks can’t see or discriminate color at all, but a recent case study reported a patient who could. It involved him needing to pick up and move an object around in front of his eyes for a bit, and then suddenly he would declare “It’s red!” You can see the videos here or read the full case report, it’s interesting. He describes it as the color popping out and stinging his eyes, and then it seems obvious.

I imagine it’s a bit like seeing an object in the dark, not quite being able to make out what it is, and then once you know what it is, it becomes obvious. It is odd to think of having this kind of perception with something as basic as color, since that usually jumps out at us so easily.

There are plenty of descriptions from blindsighted people describing their experiences. They tend to emphasize some vague awareness without quite knowing what they’re aware of, for example: “I had an impression that something was there. Where it was made a greater impression than what it was.” One individual described it as like the experience “of a normal person when, with the eyes shut, he looks out of the window and moves his hand in front of his eyes . . . like a ‘shadow’” or “a black shadow moving on a black background”.

So the picture I get of all of this is that sensing even the most basic visual elements is at the edge of their abilities, and combining them into “things” isn’t possible.

For us sighted folks, vision is very “all or nothing”. We get incredibly rich information through sight, or we close our eyes and get nothing. Part of what’s so captivating about blindsight is thinking about it as just as detailed as our normal, vivid sight, but without awareness. But the better image might be vague shadows seen through a foggy glass. Different aspects are distorted, can’t be put together into a full image, but you get occasional senses of something.

Subscribe now

Is Blindsight Just Degraded Normal Sight?

I don’t want to give the impression that all of this is settled and it’s obvious that blindsight isn’t anything special. There are ongoing debates about all this. I’ve been drawing a lot on Ian Phillips’s paper in writing this, but there are other researchers who have written responses to his arguments (to which Phillips has replied).

The gist of the debate is that, even if in some cases blindsighted folks are aware, there seem to be others where they aren’t. The debate gets technical over how well Phillips’s arguments cover all the different cases.

All that said, when I read the empirical papers and the comments debating the issue, I find myself siding much more with the deflationary accounts. Blindsight seems special in the sense that it’s very particular degradations of vision that are worth studying, but I think this tells us more about vision than awareness/consciousness. It doesn’t seem to me like anywhere near as clean a dissociation between detection and awareness as you would want if you wanted to make big claims about different conscious versus unconscious processing.

It’s interesting comparing blindsight to something like a reflex. Reflexes also involve sensing something (like a hot stove) and taking action based on it without conscious awareness. What supposedly made blindsight special is that the action taken can be flexible—you can use that discriminatory ability to take arbitrary actions like, if you see something, “hit the button” or “raise your hand”. Reflexes lack that sort of flexibility. But the above story makes it seem like, to the extent that information is there and available to be acted on flexibly, it also becomes aware.

While the popular depiction of blindsight as a clear visual discriminatory ability without awareness might make great fuel for science fiction and thought experiments, the reality is less striking. But to the extent that blindsight isn’t special, thinking about perceptions at the edge of our awareness can be a good exercise. Our awareness can be vague and uncertain, unclear whether our “eyes are playing tricks on us”. There are many shades of grey in our visual experience, including whether we’re seeing a grey circle or a faint Gabor patch.

If you enjoyed this, please hit the “Like” ❤️ button, restack, or share this article to help others find it.

If you enjoy Cognitive Wonderland, consider supporting it by becoming a paid subscriber at whatever level feels comfortable for you.

If you’re a Substack writer and have been enjoying Cognitive Wonderland, consider adding it to your recommendations. I really appreciate the support.

Subscribe now

Share

Leave a comment

If you do not want to receive posts like this about Book Club logistics (voting for books and learning what was selected), you can turn it off in your account settings:

Now on to the voting. Here are the options for July. I’m keeping the options at 3 this month, since some of these have been with us f…

Read more

Filippo Brunelleschi was a 15th century Italian architect, designer, and sculptor. He had a huge influence on Renaissance architecture, and is sometimes considered the first modern engineer. But one of his key accomplishments was in the art world: perspective painting.

Brunelleschi conducted a series of experiments to figure out the geometric rules of linear perspective. Using grids and geometric calculations of scale, he came up with a system for representing three dimensional objects on a two-dimensional surface. From one perspective, parallel lines converge to a “vanishing point“, and there are specific mathematical relationships between distance, size, and position in a visual field.

While none of Brunelleschi’s paintings survive, he influenced Renaissance painters who used and built on his technique.

This application of geometry to perspective, Wootton argues, was a key shift that led to the use of parallax in astronomy: judging the distances of celestial objects with a shift in perspective. This application allowed Tycho Brahe to argue that a supernova in 1572 and a comet observed in 1577 were beyond the moon. This meant that the world beyond the moon wasn’t unchanging, as Aristotle had thought, contradicting a key tenet in the dominant view of the universe. Math and observation gave people tools to question received wisdom and show them to be wrong. The revolution was born.

What I find interesting about this story is the connection between art and astronomy. Despite their many differences, the same geometry applied to both. The connections don’t end there, though. Similar mathematics were being used in cartography and ballistics. It was no coincidence that Brunelleschi was interested in painting, engineering, and architecture—they shared a common underlying math.

Galileo’s thinking about falling objects led to the formulation of inertia, another overturning of Aristotle’s views of nature. Newton later found that the same mathematical rules governing falling objects also held for the orbits of planets. Despite their seeming differences, math allowed Newton to see these phenomena were deeply connected: they are both the work of gravity.

Math connecting seemingly disparate things isn’t just historical, it’s shaped my personal career. As a computer science student I took a lot of math classes, and later added statistics and machine learning as part of my neuroscience training, before becoming a data scientist.

This math training allowed me to work in many different domains. It turns out the same math tools can be used to analyze the firing rates of neurons or predict the trajectory of chronic kidney disease. I’ve used math to find who is using an email service provider to send phishing emails and to detect who is falling asleep based on their breathing rate. Math has given me a depth and breadth of experiences in my professional life that would be hard to match otherwise.

Subscribe now

Math and Abstraction

Applying math to the world is a process of abstraction. We expect two balls to fall at the same speed even if they differ in some ways, like their color.

In this way, it’s not that surprising that we find mathematical connections between things. For example, similar mathematical models can be used to predict the spread of a disease or the spread of a rumor. Even though diseases and rumors are different in many ways, we can abstract away those differences and instead look at the similarities: they both spread from one person to another, and involve going from uninfected/ignorant to infected/spreading to recovered/bored. They both have dynamics where the disease or rumor either spreads out of control or dies out quickly.

The mathematics show a deep analogy between rumors and diseases.

Both analogy and mathematical modeling involve abstracting away certain details. With analogies, you peel away the parts that aren’t the same to reveal that, after all those superficial things are taken away, you’re left with the “same” thing. When I explain to my 5-year-old that gasoline is the car’s “food”, I’m making an analogy. I might say the car eats gas, just like we eat food. The analogy captures that food for humans and gas for cars are broken down for energy.

But it isn’t quite “just like” that, there are lots of differences—we are biological creatures, cars are mechanical. Combustion in a car creates mechanical force directly, whereas in us breaking down food creates ATP which can later be used to fuel cells. Some of the nutrients in food are used for growth or repair, whereas gas is just used for energy. But we abstract away these details to focus on what is similar, and it helps my son understand what cars do with gas.

In their book, Surfaces and Essences, cognitive scientists Douglas Hofstadter and Emmanuel Sander argue that analogy is the core of all thinking. This abstracting away details to decide one thing is just like another is what allows us to make categories. No two cats are exactly the same, but we abstract away the differences to create the concept cat. When we talk about a plan falling apart, we are making an analogy between the plan not working and the crumbling of a physical object, abstracting away that plans are intangible and instead looking at the deeper pattern of something losing its structure.

With mathematical modeling, we’re doing something just like that. We abstract away particulars and find some deeper structure. We find that diseases are just like rumors. We find many different measured quantities follow a normal distribution (a bell curve) because, when you strip away their differences, many processes are just like each other. The Central Limit Theorem tells us that any time you add together a bunch of independent factors, you get the normal distribution. In this way, rolling a bunch of dice and taking their sum is just like the many different genetic and nutritional factors that result in height. Both height and dice rolls form a normal distribution because of the structural relation that hides behind the many details you can abstract away.

The normal distribution shows up all over the place because it’s very common for processes to be the result of adding up a bunch of different contributing factors. Math can show us the deep similarities between things that superficially seem to have no similarities.

Subscribe now

The Poetry of Math

The scientific revolution is sometimes described as the mathematization of nature. We discovered how powerful these abstractions are for describing the natural world, and this formalization allowed us to see connections and patterns we hadn’t seen before. Our view of the universe changed. We learned the Earth goes around the sun. Falling objects, the tides, and the orbits of planets were all due to the same force: gravity. We realized the speed of light is finite, and that blood circulates the body in a closed loop, it isn’t constantly produced and consumed.

The mathematical approaches to understanding population growth by economist Malthus inspired Darwin, who saw the Malthusian pressures of population outpacing limited resources applying to all organisms. Math allowed Florence Nightingale to show soldiers were dying from preventable infections rather than battle wounds and Einstein to discover our modern understanding of space-time warping.

Math has changed how we see every aspect of the world, through showing us connections and forcing greater precision in what we mean.

And yet, many people have a distaste for math. People say they are not math people, or consider math “sterile”. We call people “calculating” as an insult. Through the process of abstracting away the details, it is sometimes argued, we take away the warm fuzzy individual facts, and leave behind the cold hard structure, reducing the individual to an equation or statistical data point.

This is an unfortunate public image problem.

It has so much to give us, but people tune out at the word “math” and brush algebra off as cold. But the process of abstracting away details to reveal surprising structure is seen as warm in poetry.

Tracy K. Smith, in her Pulitzer prize winning poetry collection, Life on Mars, compares the universe to a house party:

Text within this block will maintain its original spacing when published

A conflagration of suns? It sounds like the kind of party
Your neighbor forgets to invite you to: bass throbbing

Through walls, and everyone thudding around drunk
On the roof. We grind lenses to an impossible strength

Point them toward the future, and dream of beings
We'll welcome with indefatigable hospitality

Poets are particularly good at finding analogies that are both beautiful and surprising. They draw our attention to a surprising underlying structure, often hard to articulate in plain prose and more felt. But it gives us a new and surprising way of looking at something, allowing us to see it with fresh eyes.

I’m no poet. But math can defamiliarize things to help us see them anew just like how poetry can.

When a scientist takes repeated measurements with an instrument, small independent errors accumulate: the shaking of their hand, thermal noise in the measurement device, small fluctuations in what they’re measuring—resulting in slightly different measurements each time that form a bell curve. This is just like how your grocery bill each week varies based on small fluctuations in prices, the small additional thing you bought, or the thing you usually get that you skipped.

Like poetry, it reveals surprising connections between things, like how the geometry of art connects to our observations of planets and supernovae. And sometimes those connections lead to a fundamental shift in how we see everything.

If you enjoyed this, please hit the “Like” ❤️ button, restack, or share this article to help others find it.

If you enjoy Cognitive Wonderland, consider supporting it by becoming a paid subscriber at whatever level feels comfortable for you.

If you’re a Substack writer and have been enjoying Cognitive Wonderland, consider adding it to your recommendations. I really appreciate the support.

Subscribe now

Share

Leave a comment

We discussed the changes to the brain and hormones that accompany fatherhood, and the many challenges and benefits of fatherhood.

If you enjoyed the discussion, make sure to order Darby’s book!

What I didn’t appreciate, until I read The Mind of a Bee, is how smart individual bees are.

For one thing, I didn’t realize how sophisticated the senses of a bee are.

Their senses of smell, for example, are quite good, to the point where some researchers have tested them to detect explosives and other substances as an alternative to sniffer dogs. Aren’t airports bad enough without bees? (Insert Seinfeld riff here)

Flowers, of course, are important for a bee. They use their senses to assess flowers in surprising ways. For example, they can detect electrical charge. When bees fly, they shed some electrons to the air, becoming positively charged. Thus when they visit a flower, they take some of the flower’s electrons, giving the flower a brief positive charge. Bees can detect this charge, and don’t waste their time on those flowers likely already emptied by another bee.

Bees see in much lower resolution than we do. They have compound eyes, made out of a combination of smaller visual units (called “ommatidia”), giving their vision a sort of pixelation. Based on the number of ommatidia bees have, their vision of what’s in front of them would be a bit like seeing a 30x30 or 60x60 image.

Despite this, vision is extremely important for bees. They can navigate by landmarks and find their way back to the nest based on familiar surroundings. But more importantly, bees rely on their sight to spot flowers.

Bees burn a lot of energy. Flying around is exhausting, and the nectar of flowers is often hidden away and difficult to access. Bees need to be able to tell from afar what flowers are worth investigating.

Bee eyes are sensitive to different wavelengths of light than ours. They can’t see red, but they do see ultraviolet, and so flowers look different to them than they do to us, often revealing additional patterns we can’t see.

Surprisingly, analysis of the spread of genes through the insect kingdom suggests that bees’ color vision came first, before flowers. It wasn’t that bees developed color vision to detect flowers, but flowers became colorful to attract bees.

Without landing on them, bees can assess flowers. They learn, through watching other more experienced bees or through their own experience, what types of flowers are worth foraging from, and assess them from afar.

This social learning in particular surprised me. It’s easy to think of bees as little robots, with maybe a bit of isolated learning ability, but mostly running on hardwired rules. They live seemingly pretty simple lives, so it seems plausible that they wouldn’t have sophisticated learning abilities. It turns out that’s far from reality.

Subscribe now

Learning in bees

Social learning from other bees doesn’t stop at learning what kind of flowers to land on. Some flowers are “deep-tubed”, keeping their nectar out of reach of any short-tongued bees. Some bees learn that they can bite the base of the tube and suck the nectar out, without going the “usual” route of putting their tongue down the tube. There’s evidence that bees can learn this technique from each other, and that there’s even a sort of cultural transmission—in different areas, there are different local “traditions” of biting either the left or the right side, for example.

In the lab, a task was set up that required a bee to pull a rope to expose a “flower” filled with sugar water. Of the 100 bees tested, only 2 were able to learn this task on their own. But the other bees became able to do it after just watching another bee perform the task.

There’s no obvious “in the wild” analogue of this sort of behavior. Bees don’t pull ropes to make their hives. Their learning is surprisingly flexible.

In the lab, you can test bees by setting up artificial “flowers” and filling some of them with sugar water. Bees can quickly learn that certain characteristics of the artificial flowers get them the sugar water reward. For example, if you only put sugar water in the artificial flowers of certain colors, they’ll quickly learn which colors are rewarded—unsurprising since this is the sort of cue they would encounter out in the wild.

Other cues they can learn are more surprising. Bees can detect magnetic fields, most likely to help them navigate, but they can learn to associate magnetic stimulation with a reward. They can even learn to associate threat pheromone—normally a danger signal to other bees—with reward. This flexibility in what bees can associate with reward cuts against a view of bees as driven by simple hardwired instincts.

Bees can also associate cues with actions they need to take. They can learn that a blue entryway to a tube means they have to turn left to get to the reward, while a yellow one means they need to turn right. They can make associations not just between a color and a reward, but between a color and the rule they need to follow.

But what I found most surprising is that bees can learn abstract concepts, like “same” versus “different”. In one study, bees were trained on what’s called a “delayed match to sample task”. Bees would encounter an entryway that had a color on it, and then inside the tunnel there was a fork where the bees had to pick whether to turn left or right. The two directions they could turn were themselves labeled with a color. The experimenter could then decide on a rule—like having the bees learn that if the entry was yellow, they should take the yellow branch, a “sameness” rule, or if the entry was yellow, they should take the blue branch, a “difference” rule.

Bees were able to learn these sorts of rules. But what was more impressive is they were able to generalize it. If they were trained on finding the matching or differing color, they could then be tested on a matching or differing pattern, like horizontal versus vertical lines. Bees trained to look for the same color (yellow to yellow) would look for the same pattern (horizontal to horizontal), while those trained to look for the different color (yellow to blue) would look for the different pattern (horizontal to vertical). The bees had learned the abstract concepts of “same” and “different”.

Despite their tiny brains, bees are able to learn abstract concepts, learn from each other, and behave flexibly in a way that defies the view of them as hardwired little robots. This all seems to raise a question: if such a wide cognitive repertoire is possible in such a little brain package, why do we larger animals bother with such large brains?

If bees are so smart, why do we bother with big brains?

At multiple points in the book, Chittka makes the point that the question isn’t how bees are capable of so much with such small brains—it’s why we bother to have such large brains if small brains are so effective.

This last section is for paid subscribers, as a thank you for their support. If you would like to support Cognitive Wonderland or join our community, consider becoming a paid subscriber.

Subscribe now

Read more

The claims aren’t particularly new. In 2015, a Time article claimed “You Now Have a Shorter Attention Span Than a Goldfish”.

It’s a common belief that modern social media, short form content, and general “brain rot” of the modern age has led to the modern day populace being unable to pay attention. Yet if you talk to attention researchers, you get a different picture.

When asked if attentional skills have declined, Edward Vogel, professor of psychology and neuroscience at the University of Chicago, said: “I’ve been measuring college students for the past 20 years. It’s been remarkably stable across decades.”

The article quoting Vogen goes on:

His findings echo those of other experts, including Michael Posner, a psychologist known for identifying the brain networks underlying attention, and Marcus Raichle, a neurologist and authority on brain metabolism. They say the ability of healthy adults to pay attention hasn’t diminished.

“There is no real evidence that it’s changed since it was first reported in the late 1800s,” Dr. Posner said.

Similarly, in his book about screentime, Unlocked, psychology professor Pete Etchells asks two researchers whether they’ve seen any evidence of declining attention. Jacob Fisher says “I’ve yet to see anything that’s actually convincing.” Chris Chambers goes further:

It would be obvious if there was a decline. It would be easy to look at the last, say, fifteen years of research on attentional cueing and look at whether reorientation costs are getting higher. My virtually certain prediction is that you would get no effect whatsoever.

In fact, when I looked into it, the only study I could find that attempted what Chambers is describing found a modest increase in attention/concentration ability (more on that below).

Why might there be such a discrepancy between the general public’s perception of a decline in attention, and the researchers that study it? A clue comes from this article that quotes Gemma Briggs, yet another attention researcher who doesn’t think there’s a decline:

In fact, [attention researchers] think the idea that attention spans are getting shorter is plain wrong.

“I don’t think that’s true at all,” says Dr Gemma Briggs, a psychology lecturer at the Open University.

She studies attention in drivers and witnesses to crime and says the idea of an “average attention span” is pretty meaningless. “It’s very much task-dependent. How much attention we apply to a task will vary depending on what the task demand is.”

Briggs’s quote indicates the importance of what it is we’re doing. “Attention span” isn’t some simple quantifiable ability, it depends on “the task demand”. What it is we’re doing and the environment we’re doing it in affect how long we pay attention to it.

It’s possible we’re correctly inferring that it’s harder to pay attention for longer, but incorrect in attributing it to our diminished abilities. It might be that we have more vying for our attention.

We’ll come back to that. But first, let’s talk about the headlines pinning the blame here on our cognitive capacities being diminished due to phones and social media. With basically every news headline you see talking about a new study showing how scrambled our brains are, there are only two things you need to know to assess whether there’s anything to the study: short-term effects are not lasting deficits, and correlation is not causation.

Subscribe now

Short-Term Effects Are Not Lasting Deficits

Let’s look again at the Vice article I mentioned above: “TikToks, Shorts, and Reels Are Melting Your Attention Span, Study Finds”.

Sometimes you find articles where the headline is a bit over the top, but the substance of the article is much more even-keeled. This Vice article is not one of them. Here are a few choice quotes from the article:

It turns out “brain rot” isn’t just a cute little term to describe the current state of the internet.

Short-Form Video Is Wrecking Your Memory and Focus

short-form video, as a format, is essentially a less elegant version of the memory-wiping Neuralyzer from Men in Black

This is all bizarre if you actually read the two studies it’s based on.

The researchers had people perform a simple task: press ‘M’ on a keyboard if a nonword (like “batly”) is displayed, and press ‘N’ if a real word is displayed. Then, occasionally (about 1 in every 10 trials), the word “blue”, “purple”, or “green” would appear, and each of those words required pressing a different key (’Q’, ‘W’, and ‘E’). Participants did this for hundreds of trials, trying to respond quickly each time.

Participants did this task for a little while, and then were given a 10 minute break where they either scrolled their TikTok (or other short form video) feed, or did something else (watched a longer video, rested, scrolled Twitter, etc).

What they found is that people who viewed TikTok during the break tended to perform worse on the occasional color word test (but did fine on the real word versus nonword test). They put this down to the rapid context switching involved in short form videos disrupting “prospective memory”. That just means they forgot that they were supposed to do something different for color words, or forgot which button they were supposed to press for which color.

This doesn’t seem terribly surprising to me. Rapid context switching makes it hard to remember things. I experience this every morning when I’m trying to get my kids’ stuff ready for preschool and they stop me every 30 seconds to ask me to get them milk or to look at how high they can jump. When I return to getting their lunches together, I’ve totally forgotten what I was going to do next in the process.

Not to say there’s nothing we can take away from this research. It’s probably useful to know that, if you’re taking a break from work, scrolling TikTok might not be your best choice. You might go back to work more scattered than you would be if you took a walk, talked to someone, or watched one thing instead of 20 different short videos. But this is completely different from showing some form of lasting cognitive harm.

I’m going through these studies as one example, but this type of research—showing exposure to screens/social media/short form videos has some immediate deleterious effect on some cognitive skill—is common. The experiments are easy to run and they tend to get a ton of media attention. Another recent one had people watch TikTok videos and then read a news article, and found people who watched TikTok paid less attention to the news article. Watching short term videos may have a bit of inertia, putting you in a state where there’s a mismatch with the slower pace of reading, but again, this is very different from a permanent impact on cognition.

This stuff is useful to know. If you are in the middle of doing something hard, and take a break with TikTok, there’s a good chance it’ll be hard for you to get back on track afterwards. But that isn’t evidence that using TikTok negatively impacts your ability to pay attention in a lasting way.

Subscribe now

Correlation is not Causation

“Correlation is not causation” has become a cliche, but it seems it needs to be said more often because the media has not learned it.

I would wager that most headlines you read claiming that social media is reducing our ability to pay attention are conflating correlation and causation. Typically if you look at the studies, they will say there is an “association” or “correlation”.

Take this one: “TikTok scrolling can cause ‘brain rot’ according to new American Psychological Association study”. Notice the explicit use of the word “cause”. This is in direct contradiction of the academic paper itself, which says “most included studies were cross-sectional and correlational, the findings of this review cannot confirm directionality”. In other words, it’s correlation, not causation.

There’s similarly been much ado about studies showing an association between brain volume and short form videos. Again, these are correlational. In the words of psychology professor Pete Etchell:

“[From this study] you can’t say anything about whether watching short videos causes changes in the brain, or whether certain types of brain structure precede certain types of video consumption.”

This is the key problem with interpreting any correlation. Are those with pre-existing attentional problems more drawn to TikTok? Probably. Are there brain differences between those with attentional problems and those without? Probably. So finding a correlation between social media usage and attentional problems or brain differences doesn’t tell you the social media usage caused the problems.

Of course, something doesn’t need to establish causality to cause us worry. But when we take a step back and look at the broader picture, I think it’s fair to say the concerns are overblown. In general, the association (again, correlation, not causation) between social media usage and cognitive effects are pretty small and inconsistent—certainly not large enough to justify the “brain rot” rhetoric you frequently see.

Combined with attention researchers saying measures of attentional ability have remained stable, there isn’t much reason to worry. But that hasn’t stopped certain sensational claims about our vanishing attention spans.

Subscribe now

Are Our Attention Spans Shorter Than Goldfish?

About ten years ago, a slew of articles from major outlets—Time, The Guardian, Telegraph, USA Today, National Post, and the New York Times—claimed that our attention spans have shrunken to 8 seconds long (shorter, it is claimed, than a goldfish).

In this case, it isn’t an exaggeration of some published research. It’s actually not clear where the number comes from. Journalist Simon Maybin traced it to a report from a Microsoft team, which cites a company called Statistic Brain, which cites the US National Library of Medicine and the Associated Press. When contacted, none of these organizations could find any record of research that backs up those numbers. It’s totally unclear where that number comes from.

There is, however, some real research that seems to tell a similar story. Gloria Mark has done studies looking at how long people stay on one task in the workplace. She performed a few different studies over the years, and makes statements about how in 2003 the attention span was 2.5 minutes, in 2012 it was 75 seconds, and in 2016 it was around 40 seconds (and seems to have leveled off).

She is talking about three studies, and there isn’t a problem with any individual one. But the problem is comparing across them. For one thing, they use radically different measures—in the 2003 study, a person sat with a stopwatch and tracked when a person “changed task”, but the 2012 and 2016 studies use a computer logging measure that tracks when someone changes the window currently in focus.

As I write this, I am switching back and forth frequently between multiple windows. I have one window where I am typing up this document, and another with a few tabs with some of the research I’m talking about. I’m flipping back and forth between them whenever I reach a point where I need to consult a detail in the research. This is similar to when I’m programming, and rapidly switch back and forth between documentation, a window testing what I’m coding, and the window where I’m writing the code. Each switch would be considered task switching by the measure of Mark’s later studies, even though I’m on the same task with each switch.

In the first study done with a stopwatch, all of these window switches may have been considered the same task. Or maybe computer window switching has become more common as the nature of work has changed over time, and now often involves multiple applications.

These studies were also done in different companies with different people (they’re all small samples, with 13, 14, and 40 people in the three studies), so the differences might also just be due to differences in the workflows in those companies or with those people.

The point is, comparing the numbers just isn’t meaningful. We don’t know what’s caused it, and it certainly doesn’t say anything about attention spans. But that hasn’t stopped Mark from making sensationalized claims and writing a book about it.

Subscribe now

The Case Against Rotting Brains

The problem with writing an article like this, where there’s a bunch of stuff to debunk, is that I’ve spent a huge chunk of this article giving examples that go against my thesis (which is that attention spans haven’t shrunk) and arguing it’s weak evidence. You might agree no individual study is perfect, but still think the overall weight of evidence—and your experiences from your own life—is in favor of shortened attention spans.

I often hear about other supposed indications that our attention spans have dwindled—fewer people reading for pleasure, or the fact that people spend time watching short videos on TikTok itself as an indication of our reduced ability to pay attention. But we can also point to other forms of sustained attention—movies have gotten longer, and binge-watching shows has become commonplace. Long “video essays” on Youtube and other streaming platforms are increasingly popular among Gen Z. If you go looking for indications in either direction, it’s easy for confirmation bias to serve up a few examples where we spend more or less time paying attention to specific things.

It’s much better to look at actual studies instead of trying to draw inferences from trends in media. The technology and media landscape have changed. We should be looking for some longitudinal change in something that actually measures our ability to pay attention.

This takes us back to the quotes from various attention researchers we started with, saying they haven’t seen a change in attention measures in the decades they’ve been studying it and would be very surprised to see some.

The only study that I’ve found that directly compares attentional ability across time actually found that concentration performance in adults has increased from 1990-2021. They call this the “Flynn effect for attention” — the Flynn effect being the noted increase in IQ across time. People are getting smarter, and apparently, getting better at concentrating.

Now, to be clear, that is one review, looking at one particular operationalization of attention (the d2 test of attention). I don’t think we should go around claiming attention has been getting better based on this single study, but I do think it represents the best test of what we actually want to know when we ask if our ability to pay attention has changed over time.

Given how easy it would be to find longitudinal differences, and how eager people are to find declines that all the big news organizations will parrot made-up numbers without checking them (as shown by the “goldfish” stat mentioned above), the fact that we don’t have anything close to a knock-down study says something. I think we can be pretty confident that our ability to pay attention has not decreased.

Subscribe now

Why Does It Feel Like Attention Spans Are Shortening?

As mentioned in the intro, about half of people think their attention has gotten shorter. The term “brain rot” has gained popularity for a reason. There is certainly a perception that phones have wrecked our brains and made it harder to concentrate.

But I think there’s a simpler, more likely explanation that makes sense of this association that many people have about phones taking away our ability to concentrate. Any moment of boredom I feel, I can whip out my phone and start scrolling social media. There’s no friction. 25 years ago, this wasn’t an option.

Phones make attention-wandering both easier and more visible than they used to be. If you’re at dinner and your partner pulls out their phone, you know they aren’t paying attention in a way that’s much easier to tell than if their mind just started wandering. Almost every time I procrastinate, it’s by looking at my phone. It would be easy for me to conclude that the reason I have trouble doing anything hard is because my phone has reduced my ability to concentrate on what I’m doing.

I certainly don’t want anyone to take this article to mean “no one actually has a problem with phones or social media”. It’s worth being mindful about what role these apps play in our lives, and if we’re having trouble with procrastination because of them, we can take steps to mitigate that. But there’s a big difference between a drop in our cognitive abilities and there being something distracting in our pocket.

We have control over our environment, especially over our phones. We can use them mindfully, gaining the things we like about phones (keeping connected socially, staying informed about things we care about, easy entertainment when waiting at the doctor). Or you can leave all notifications on and let it distract you every 5 minutes while you’re trying to get work done.

We have control here. Our brains haven’t rotted, and that’s good news. If you feel like you’re easily distracted and want to concentrate more, there’s no need to go to the extreme of swearing off all phone use. Just add friction to using it when you want to concentrate. Put it in another room, put it on Do Not Disturb mode, install an app that helps you manage usage.

You have control, your brain isn’t rotted.

If you enjoyed this, please hit the “Like” ❤️ button, restack, or share this article to help others find it.

If you enjoy Cognitive Wonderland, consider supporting it by becoming a paid subscriber at whatever level feels comfortable for you.

If you’re a Substack writer and have been enjoying Cognitive Wonderland, consider adding it to your recommendations. I really appreciate the support.

Leave a comment

Share

Subscribe now

• From the back cover: The science of emotion is in the midst of a revolution on par with the discovery of relativity in physics and natural selection in biology. Leading the charge is psychologist and neuroscientist Lisa Feldman Barrett, whos…

Read more

Subscribe now

We know that babies’ vision takes some time to develop, but by 6 months, they are seeing with pretty much adult-level clarity. And yet studies show young kids suck at looking for stuff long after infancy.

If kids are getting the same clear image as the rest of us, and they want to find the Easter eggs, why can’t they find them? The problem isn’t their vision, it’s their visual attention, which continues to develop long after infancy.

It’s common to think of it as a sort of spotlight that we can sweep around, where we direct the spotlight where to go. But what does that mean for what’s going on with my 2-year-old? He wants the Easter eggs, he’s looking in the right direction, and his vision is just as good as mine (maybe better). Yet something isn’t quite working yet. What is that something? Why isn’t his “spotlight” falling on the egg?

Attention is one of the most intimate aspects of our mental lives. Where our attention goes, we go. Shifting attention feels like a pure act of will, not like the sort of thing with biological mechanisms that can be immature. But to explain what it is that kids are lacking, we have to investigate attentional mechanisms. It might not feel like the kind of thing that can be broken down, but it quite literally can.

Subscribe now

Hemispatial Neglect

Hemispatial neglect is a disorder of attention. It generally affects people who have had a stroke damaging areas (most commonly the parietal cortex) of one hemisphere, and makes them weirdly inattentive towards anything to one side. They may only eat food on the right side of the plate, only dress the right side of their body, only shave the right side of their face, and so forth.

This isn’t an issue with their vision. They can still see on their neglected side, their attention just doesn’t go there. If you have them draw something, they’ll tend to miss features on the neglected side.

Notice that the neglect isn’t complete, they’ll still draw some stuff on the neglected side. Careful experiments show that people with hemispatial neglect actually can have their attention go to their neglected side, but only if there isn’t anything in their non-neglected side. As soon as there is an object competing for visual attention from their favored side, whatever is in their neglected side seems to just vanish from awareness.

It’s like there’s a competition going on, and the neglected side will lose if there is any competition from the other side.

Subscribe now

Biased Competition Theory

When we’re looking around in a complex, cluttered environment, there are many different objects with different features, any of which could capture our attention. We can’t be aware of everything all at once, so what determines what “pops out”?

The way attention researchers generally think about how attention is focused is by two factors: top-down and bottom-up.1

Bottom-up factors are simple: think about anything we consider “eye-grabbing”: shiny, bright, colorful, and so on. Some simple visual characteristic makes it “pop” and draws our eye.

Top-down are factors based on our current goals and priorities. If you’re searching for something yellow and egg-shaped, you’ll notice more things with those features.

These factors combine to create a sort of competition for our attention. The neural representations compete across stages in visual processing, inhibiting each other to become the dominant signal. At different stages in this visual processing, our top-down goals affect the competition: projections from the frontoparietal cortex modulate the gain to amplify signals related to our current goals.

The theory describing this is dubbed “Biased Competition”, since the idea is the low-level features are competing while the high-level goals bias the competition. And while I’ve been describing this all in terms of visual attention, it’s worth noting the same basic competition framework holds for other forms of attention.

With young children, this top-down biasing hasn’t fully developed yet. Their visual attention is more drawn by bottom-up features. Not completely—even infants do fairly sophisticated goal-directed looking—but more so than adults.

So when my 2-year-old is searching around a space completely covered in brightly colored toys, it’s no surprise spotting the brightly-colored egg is hard. He can’t bias the competition in favor of egg-shaped objects as effectively as I can, so the eggs are less likely to “pop out” for him.

As children grow, they are gaining new abilities to control themselves and their impulses. They are gaining abilities to pursue their goals more (as well as the ability to form longer term goals as their memory and understanding of the world develops).

In people with hemispatial neglect, there’s been damage to this frontoparietal network that does the biasing. Normally the two hemispheres are each biasing things in their direction and inhibiting the other hemisphere, sort of like how in a tensed arm the bicep and tricep are active and balancing each other out. Damage to one hemisphere releases the other from that competition, so attention is consistently pulled in that direction. The neglected side can win if nothing on the other side is competing, but as soon as there is serious competition, it loses.

Subscribe now

Why Doesn’t Attention Feel Like a Competition?

I suspect the spotlight metaphor feels apt because it goes along with what it feels like to direct our attention. It doesn’t generally feel like there’s some sort of tug of war happening within us, with different things competing for our attention. The things we’re not paying attention to, we’re simply not aware of. This goes deeper than you might think. Patients with hemispatial neglect frequently also have anosognosia: a lack of awareness of their impairment. Despite the fact that hemispatial neglect can be extremely debilitating, people who suffer from it frequently don’t notice it. The same systems that would alert them that something is missing are the ones that are damaged. If even catastrophic failures of attention can pass by without us being aware, it’s no surprise we aren’t aware of all of the behind-the-scenes workings of attention. Awareness itself is something that needs to happen explicitly, it doesn’t happen for free just because something is happening inside of us.

This is part of the awkwardness of explaining, in third person terms, what’s going on when we do something. Part of this isn’t particular to attention, or even to cognition. When you contract a muscle, there is a long chain of biochemical events that happen, beginning with the electrical activity in a nerve and culminating in a series of “power strokes” of myosin, which repeatedly binds to, moves, and disengages from the muscle fiber. We don’t feel these series of movements, we just feel the muscle change shape and tense.

Similarly, we could describe the logic gates at work in your computer while your computer runs a game like Doom, but it would feel completely different from the level of description involving characters on the screen. We could describe the fluidity of water or the stickiness of tape in terms of chemical properties, but it would (at least initially) feel completely different and disconnected from the higher level phenomenon we’re used to.

It takes deeply reconceptualizing something to grok what it means to explain a phenomenon we’re used to interacting with at one level at another level.

But there is something a bit special about attention and other cognitive functions: they’re a description of a part of us, the very thing doing the understanding. Seeing ourselves described in third-person ways can feel like it takes away agency, like these are things that just happen without us, as if we’re being taken out of the picture of the functioning of our own minds.

Describing attention as a competition between neural representations, with gain modulation from frontoparietal networks—this seems to lack our role in the process. But we know from our experience, we can decide to direct our attention somewhere.

If you decided, in an act of rebellion, to show this picture is wrong by directing your attention at something arbitrary, that top-down goal affects your attention.

It might feel like there’s a tension there, but there isn’t really. The whole point is that these mechanisms are part of who we are. Your goals are the top-down bias. When you want to find Easter eggs, it’s that desire that influences what pops out at you. It isn’t something happening to you, it’s you, your desire, that is what does the biasing.

Understanding this is what it means to better understand ourselves. We aren’t black boxes, we can look at the mind and ask how different aspects work and what their purpose is, and through this view ourselves with a different lens.

The different aspects of our mind are what allow us to live our lives. The competition that is attention is what allows us, as beings navigating a chaotic world, to sort out what matters and act on it.

If you enjoyed this, please hit the “Like” ❤️ button, restack, or share this article to help others find it.

If you enjoy Cognitive Wonderland, consider supporting it by becoming a paid subscriber at whatever level feels comfortable for you.

If you’re a Substack writer and have been enjoying Cognitive Wonderland, consider adding it to your recommendations. I really appreciate the support.

Leave a comment

Share

Subscribe now

If you do not want to receive posts like this about Book Club logistics (voting for books and learning what was selected), you can turn it off in your account settings:

Now on to the voting. Here are the options for May. I’m carrying three forward from the last poll and one new one at the bottom:

• The B…

Read more

Take for example the story of The Stonecutter (what follows is an abbreviated version):

A poor stonecutter worked every day harvesting slabs from a great rock. One day he delivered a slab to a rich man, and envied the rich man’s possessions and leisurely life. He wished he were a rich man.

A voice answered: “Your wish is heard; a rich man you shall be!”

As a rich man, the umbrella his servants held over his head didn’t keep the sun’s rays from burning his face. He cried out: “The sun is mightier than I; oh, if I were only the sun!”

And a voice answered: “Your wish is heard; the sun you shall be.”

As the sun, a cloud hid the earth from him. “The cloud is mightier than I? Oh, that I were a cloud!”

And a voice answered: “Your wish is heard; a cloud you shall be!”

As a cloud, he rained down on the earth. But the rock was not affected by the rain. “Is the rock, then, mightier than I? Oh, if I were only the rock!”

And a voice answered: “Your wish is heard; the rock you shall be!”

As a rock, he was not affected by the heat of the sun or the force of the rain. But along came a stonecutter who broke a great block off his surface. “Is a mere child of earth mightier than a rock? Oh, if I were only a man!”

And a voice answered: “Your wish is heard. A man once more you shall be!”

And a man he was. His bed was hard and his food scanty, but he had learned to be satisfied with it. And as he never asked for things he did not have, or desired to be greater and mightier than other people, he was happy at last.

The story begins and ends with the stonecutter in the same place: as a humble laborer. And yet in another sense, there’s been an enormous shift: he’s happy at last. The loop hasn’t changed his material conditions, but it caused a shift on a different level—a shift in perspective.

In the immortal words of Terry Pratchett:

Why do you go away? So that you can come back. So that you can see the place you came from with new eyes and extra colors. And the people there see you differently, too. Coming back to where you started is not the same as never leaving.

Subscribe now

Moving by returning where you started

Sound has a ceiling. We can’t hear things higher pitched than about 20,000 Hz. This puts a constraint on how high pitched a song can get. You can’t have a song that simply goes up in pitch forever.

Or can you?

The Shepard–Risset glissando is an auditory trick where two tones are played at the same time one octave apart. The lower one gradually gets louder as the higher one gets softer, until it finally fades out, to be replaced with a soft low note. The result is something that seems to just keep increasing in pitch.

Using a similar idea, you can create a Risset rhythm that seems to infinitely increase in tempo.

In both these cases, we have a loop — the music loops around to the same pitch/tempo, and yet it feels higher or faster because there was no point where it stepped back down. We’ve looped around, and yet there’s been a psychological change in the listener: they feel like they’ve been going “up” for the duration of the loop.

In one sense, there is movement with these loops—inside the listener. In another sense, there is no movement, just an endless circle. And yet the loop never really goes anywhere. Eventually, as I’m listening to these and my attention wanders, the effect falters for a moment and it seems to “reset”. And then I simply get bored and stop listening to it.

Any loop that doesn’t come back precisely to where it began eventually ends.

Subscribe now

An Imperfect Circle

There are a number of species of ants considered “army ants”. They form enormous swarms to forage. They raid wasp nests, hunt spiders, and sometimes even eat vertebrates like lizards, frogs, or birds. The colonies are typically nomadic, never staying in one place for long because they quickly exhaust their environment to feed their enormous colony population.

This specialized social foraging requires coordination. To stay together, the ants lay down pheromone trails as they travel, and they return to their temporary nesting site by following the same pheromone track back.

Occasionally, a group of ants can get lost from the pheromone track. Each ant will simply follow the pheromones of the ant in front of it, but if the lead ants encounter an obstacle and need to turn back, they end up following the pheromone trail of those following it. The result is a circle of ants, each following and reinforcing a pheromone trail that leads in a circle. This is called an ant mill.

These loops aren’t perfect. They don’t go back to exactly the same state they were before for a simple reason: ants are living beings. With each loop, they are slowly depleting their energy. Sadly, ant mills generally end with the ants involved simply dying of exhaustion.

Self-Reference and Reproduction

Ant mills are loops that kill. Their imperfection subtracts, slowly draining ants of energy while not leading anywhere else. Yet there are much more persistent loops involved in life.

My son has lately been learning all about the life cycles of animals in preschool. Egg to caterpillar to chrysalis to butterfly, which lays eggs to begin the cycle anew. I’m not sure he totally gets it—he likes to pretend to be a caterpillar who turns into a butterfly, but then as a butterfly he’ll turn back into a caterpillar. I haven’t had the heart to tell him that the way the cycle continues is that the butterfly lays eggs and then dies.

This cycle is possible because we animals contain the instructions for building ourselves within us. We’re all used to this idea, but it’s worth thinking about it for a moment, because it’s pretty bizarre.

Imagine trying to make a machine that creates itself. You might imagine a 3D printer that can print off all of the parts of itself, but that doesn’t quite work—it might make the parts, but it also needs to assemble them, wiring and all.

It’s a bit easier to create a computer program that creates a copy of itself. A quine1 is a computer program that takes no input, but outputs itself.

If you know anything about programming, this might initially seem puzzling—to output, a program would need to use a function to output, usually something like “print”. So to “print” itself, it would need to start with “print” and then contain the entirety of the code itself. But the code itself contains “print”. So you might try:

print “print”

Where the bit in the quotation marks is what gets printed. But notice now we’re just printing “print”, and missing the second print in the quotation marks. So to print out the previous code, we would need something like:

print “print “print””

But now we’ve added something to the code, and once again, we’re not printing the full code. We can keep going like this, adding one more “print” each time, but all we’re doing is printing out the code of the previous version. If we added a bunch of prints and ran the code, and ran the output, and so on, we would run ourselves down just like an ant mill.

We need to pull a little trick, adding a self-reference to the code to keep this loop from going on forever. The trick is that the program stores its own code as data, a string it can print literally and use as instructions. The same text ends up doing double duty. (See footnote 2 for an example solution)2

So in a way, quines “self-replicate”. If you pipe their output into a file, they create another version of the exact same code that you just executed. But that’s not very useful.

Quines are mostly just a novelty for nerds. They’re an odd puzzle that it isn’t immediately obvious how to solve, yet they’re possible in all Turing-complete programming languages (basically any real programming language).

More interesting are theoretical von Neumann probes: space probes that can create more of themselves using materials widely available in the galaxy. Imagine sending off a probe to a distant solar system, the probe autonomously collecting materials from asteroids or planets there and using those materials to create more probes, and then those probes flying off to other solar systems.

This is an idea explored in various ways in science fiction, often with these self-replicators getting out of control, becoming a cosmic virus running rampant in a galaxy with no immune system, mindlessly consuming everything to create more of themselves.

In other words, becoming like life.

Subscribe now

Life

If one were to make a self-replicating probe that frequently made errors in the copying process, it wouldn’t proliferate. With frequent enough errors and short enough probe longevity, it would just create a bunch of non-functional copies and then stop working instead of overrunning the galaxy. That doesn’t make for a good science fiction story.

This is the problem that early organic molecules had to “solve” to produce life. We’re used to thinking of evolution as powerful due to the change that comes in from one generation to the next, but that’s secondary to the primary ingredients needed for life to get off the ground in the first place: stability and fidelity.

The universe is populated by stable things. There are a lot of atoms around because they are stable configurations for subatomic particles. The molecules we come across exist because they are stable. Anything that wasn’t stable wouldn’t stay around long enough to become common.

For life to get started, there first had to be molecules that interacted with their environment in such a way that they created more copies of themselves. Common molecules are stable configurations of molecules. Richard Dawkins has called “survival of the fittest” just a special case of the more general principle, “the survival of the stable”.

Ironically, it was the lack of stability in the copying process that led to finding greater stability. Small copying “errors” led to configurations with greater stability or copying fidelity, and those overtook the replicators that came before.

It’s thought that before DNA, RNA was life’s primary information storage molecule. The switch to DNA was because of DNA’s greater chemical stability.

The great irony, then, is that the instability of the previous iterations is what leads to evolution finding greater stability.

Which came first, the chicken or the egg?

This old riddle is seductive because it seems impregnable. It evokes the idea of a perfect, eternal cycle. All chickens come from eggs, and all eggs come from chickens. How could this cycle get off the ground to begin with?

Of course now we know the answer: evolution. Through the gradual accumulation of small changes, each subsequent generation became a bit more chicken-like, until finally a generation came that we would say are chickens. Imperfect loops built chickens.

Quines, on the other hand, have the problem that their copying fidelity is too perfect. Run it as many times as you like, it will just make the same thing. Nothing accumulates, nothing changes. The loop just goes round and round, never really doing anything. Yet replication in the messy real world has led to everything from the incredible branching structures of trees to the complexities of the human brain.

Life is full of loops other than the life cycle. We wake up in the morning, we go to sleep at night, only to wake up again in the morning. We get hungry, obtain food, and eat, so that we have energy to find more food. Put in such stark terms, these loops feel pointless, like they don’t go anywhere. Perhaps it’s put most starkly in the lyrics of the band Metric:

Buy this car to drive to work

Drive to work to pay for this car

But each of these cycles is imperfect—they don’t lead to the exact same state that we were in before. And it’s a good thing, too. We don’t want to be quines, perfectly cycling back to where we were before.

In between the time we wake up and sleep (or between driving to work and making car payments), we do things that add to our life stories. Maybe we go to bed having had a new experience, or learned something, or with one more small piece of our life story in place. Maybe we’ve helped someone, affecting another person’s loop in some small way.

Like the ant mills, eventually the cycles end because we just wear out. But unlike ant mills, each cycle has meaning, and the limited number of repetitions we make around the loop is what gives the whole process value. If our time was infinite, how we choose to spend it wouldn’t have meaning.

We move our legs in a cycle to walk—our muscles and joints return to roughly the same position with each cycle. But each step brings us somewhere. Just like each sweep of your eyes across this screen, you see new words that hopefully add up to something, despite your eyes returning to the same left margin time and time again.

When you finish reading and return to whatever you were doing before, you’ll be much the same person you were when you started. But, perhaps this little cycle has made some small change in you—perhaps not as profound as the change in the stonecutter who found peace in his circumstances, but a change nonetheless.

(If you enjoyed thinking about these loops and self-references, check out Gödel, Escher, Bach by Douglas Hofstadter. Reading it nearly two decades ago inspired this post).

If you enjoyed this, please hit the “Like” ❤️ button, restack, or share this article to help others find it.

If you enjoy Cognitive Wonderland, consider supporting it by becoming a paid subscriber at whatever level feels comfortable for you.

If you’re a Substack writer and have been enjoying Cognitive Wonderland, consider adding it to your recommendations. I really appreciate the support.

Leave a comment

Share

Subscribe now

Can a mind understand itself? Perhaps, but it might take some contortions. Luckily, minds are flexible.

The purpose of The Mind’s I is to shake up our thinking about the mind, using stories and thought experiments to “provoke, disturb, and befuddle its readers”. The hope is that this will help shake off bad habits of thinking about the mind.

What sorts of bad habits? One example is the idea that consciousness is an extra added ingredient. They point to science fiction stories of conscious robots, where part of the process for creating them involves adding a self-concept and consciousness, similar to how one would add an egg to dough.

But if consciousness is something that can be added after the fact, on top of all of the functions the robot is able to perform, could it be removed? One story in the book explores this: A man lives an unbearably painful life and wishes the pain would end. He hears of a drug that can eliminate the mind, but leave the body functioning exactly as it was before. This way, he could end his pain but not harm his friends and family that would be sad to lose him. No more mental experience of pain!

He goes to bed, planning in the morning to wake up and go get the drug. But, while he sleeps, a friend who knows of his plight sneaks in and administers the drug, hoping to help him.

When the protagonist awakes… He acts exactly like he would have before the drug was administered. He goes and obtains the drug and injects it into himself. He waits a short period for some effect, and, since it has no impact on his behavior, he loudly declares “It didn’t work!” He continues complaining about his pain and trying to think of a way out of his plight.

Does this little story prove consciousness isn’t an add-on, separate from function? No. But it does help one appreciate how odd a thing it would be if it were, and invites us to rethink the connection between conscious experience, beliefs, and actions. And that’s the point—playful exploration of implications rather than clean resolutions.

Another simple thought we sometimes have is that our minds are completely transparent to us—that we, and only we, know exactly what our experiences are like. Yet, we can imagine two long-time coffee drinkers, Mr. Chase and Mr. Sanborn, who both loved Maxwell House coffee years ago and now do not. Chase says the coffee tastes like it always did, but his preferences have changed. Sanborn, on the other hand, says his preferences are the same, but the flavor itself is different—perhaps something with his tastebuds has changed.

Chase and Sanborn seem to be describing two different situations, but is it possible they could be wrong? Maybe Sanborn is failing to notice how over the years his palate has gradually been refined, or Chase hasn’t noticed how years of smoking have gradually changed his ability to taste.

We can imagine third-person investigations that would help differentiate these two cases. We could perform intensive investigations looking at the physiological responses of their tastebuds (and olfactory receptors, since smell is a large part of taste), and compare that to the responses from years ago. This could inform us whether the tasting apparatus itself has changed.

We could perform additional tests on discriminability. If the taste buds were the same, and yet we found a greater ability to discriminate different coffee varieties, this suggests the change that’s taken place is “further upstream” from the taste buds. This would be evidence that the input has stayed the same, and yet the signal is processed differently.

The upshot of all of this is these third-person investigations could tell us something about the experience and what’s changed. It seems odd, but a third-party investigation might be better positioned to tell us about our experience than we ourselves are. Perhaps, then, we don’t know as much about what it’s like to be us as we think.

Understanding our own minds is hard, but thinking about other minds raises its own strange questions.

Subscribe now

What is it like to be a BAT?

Philosophy of mind has always had flights of fancy—one of the classic papers is Thomas Nagel’s “What is it like to be a bat?” Nagel asked this because bats are creatures so different from us. They echolocate, fly around, hang upside down, and seek out bugs. But they also are conscious, intelligent creatures. We can imagine there’s something it is like to be them—Hofstadter playfully suggests they fall in the category of “be-able things”, or BATs.

What is it like to be a molecule? A windchime? A hippocampus? While we can ask, it isn’t clear these questions make sense. Surely there’s something it is like to be another sentient being in an important way that there isn’t a way to be a windchime. A windchime isn’t a BAT.

Is there something it would be like for you to be me? If anything is a be-able thing, another human must qualify. But if you were me, you would be you being me, not me being me. There’s something a bit fishy here—does it really make sense to ask what it would be like for one thing “to be” another? Grammatically we can ask “What would it be like for my ukelele to be my guitar?”, but it’s unclear what that sentence means. I guess it would be like if my ukelele… got a longer neck, a couple of additional strings, changed the coloring and markings to just… be my guitar. But then it wouldn’t be my ukelele anymore, it would just be my guitar.

What you really want to know is what it’s like for me to be me. But there’s only one person who can be me being me, and that’s me.

But you can imagine what it would be like to be me. I can tell you about an experience I’ve had, and you can imagine how I felt. You can “put yourself in my shoes”. You can ask me how I’m doing and I can describe it, in a single word (”Good”) or by describing all the physical sensations, thoughts, and feelings I’m having. Our shared experience of the human condition allows you to imagine what it is like to be me in all kinds of ways.

You could put me in a brain scanner and, with enough neuroscience knowledge, learn all that stuff about what I’m experiencing. Similarly, we could study bats and know all sorts of things about what they’re aware of, whether they’re hungry, and so on. There are plenty of third-person, objective facts telling us about “what it is like” to be a bat. We could then imagine ourselves having some of those experiences or being aware of those things.

But bats are different enough from us that our powers of empathy start to strain. It’s harder to put ourselves in their shoes.

The more different the creature, the more empathy strains and we lose our grip on what we’re trying to describe. We seem to want to know objectively what it would subjectively be like to be something. But the nature of being something means you can’t escape being it and just try being something else. We are humans, not bats. You are you, not me.

When we wonder what it’s like to be something else, and worry that any description would leave something essential out, what we’re really expressing isn’t a desire for third-person information. We want to have the experience. We can simulate the experience somewhat by mapping some of their experience onto our own to get some taste of it. Our “being” overlaps, and we can keep some parts of ourselves constant and just imagine changing some bits to put ourselves in someone else’s shoes. But the more bits we need to change, the less coherent the question becomes. We’re stuck with only being able to be ourselves.

Subscribe now

The Turing Test

The Imitation Game, now known as the Turing Test, was proposed by Alan Turing in 1950. Turing thought the question “Can a machine think?” was too ambiguous to get a clear answer, so proposed a different question: “Can a machine imitate a human in conversation?” To answer this question, Turing proposed having a judge chat with a human and a bot, and see if they can tell who is who. Can the bot imitate a human well enough to trick the judge?

The Mind’s I contains the original paper where Turing introduced the test, and also commentary on it. It’s interesting looking back on that now, in the age of Large Language Models, more than 70 years after Turing published his test and nearly 50 years after The Mind’s I was published.

Read more

At the time, I was craving some sort of creative output. I had done some fiction writing, but never a serious attempt at writing non-fiction for a general audience. I had ideas I wanted to explore and no outlet for them.

So I started writing. To my surprise, 2 years later, I’ve managed to put out something every week since. In an even bigger surprise, I now have a substantial audience.

Here are a few reflections on this little journey I’ve taken.

Subscribe now

Writing for a general audience isn’t “dumbing down”

There’s this belief some people have that to make your writing accessible and reach a big audience, you have to “dumb it down” by using smaller words and less complex sentences. You can find tools online for assessing the reading level of your text and trying to make it simpler.

To me, this is a pretty gross way of thinking about writing (and communication more generally). It’s holding the readers in some kind of contempt. I’m not some higher being trying to spoonfeed my fantastic knowledge to the barely literate unwashed masses. I’m mostly just a guy interested in stuff, trying to write about it in a way that convinces me that I understand it.

Instead of trying to use little, simple words, I think about conceptual connections. “Connections” is a pretty big word! But I can be pretty sure most people have encountered it, and will get the right “vibe” from the word.

Like most people, I have a bunch of concepts that most others don’t. We all have some shared concepts and some that are more unique to us and our idiosyncratic backgrounds. If I was discussing whether a movie was good, and you brought up that it had made a lot of money, depending on who I was talking to I might say “That’s orthogonal to whether it was good!” which is a sort of nerd-talk for “making money and being good are two unrelated things”. But I’m self-aware enough to use the non-nerd version unless I know I’m talking to someone who thinks in math concepts.

I’ve found the challenge with “translating” back to a more commonly held concept from one that is more idiosyncratic comes from my own limitations. I shouldn’t blame others for not having a particular concept I do, but I should blame myself for not having rich enough connections with that concept that I can translate it into shared concepts while holding the integrity of the idea together.

Having more richly connected concepts—meaning really understanding what you’re saying—makes it easier to find another way to say it. When I find it hard to find non-technical words to say what I want to say, that’s usually a sign I don’t really understand it that well.

Writing using more common terminology and concepts doesn’t make you look or feel smart. But I think it’s good exercise to try to put what you’re thinking into different terms. If you only write about science for other scientists, it’s easy to get carried away with the technical terminology to the point where it’s hard to describe in any other way and it loses touch with why the work matters.

Science and philosophy are both, in their own ways, relevant and meaningful to our lives and place in the universe. Finding how to connect them to our shared concepts as humans is an important exercise in keeping them connected to that meaning.

Often what’s being described in technical jargon is actually pretty simple. Take a look at this lovely example from Chenchen Li translating a paragraph from the introduction of a science paper. Here’s the original:

Animals rely on visuomotor transformations to convert object locations in eye coordinates into directional movements. The brain regions and neural circuits regulating this transformation in both vertebrates and invertebrates have been characterized. The precise neuronal connectivity underlying learned and innate visuomotor tasks is shaped by genetically hardwired mechanisms, experience or both.

Dombrovsky et al. 2025, Nature

Here’s Chenchen Li’s “translation”:

Sometimes I have to decide how to move based on how something else I’m watching moves. This means converting eye coordinate information to my own movements. We know where in the brain this happens for vertebrates and invertebrates. The exact neural connections involved are either nature, nurture, or both.

Chenchen Li’s version

There are big differences in the “vibe” of the two paragraphs. The more technical one uses certain concepts (like visuomotor transformations) that become entities of their own that map easily onto work other scientists have done. Scientists use this sort of terminology to carve out and specify a place we need a mechanistic understanding of what’s going on.

But other than explicitly trying to call to mind these technical concepts, they say the same thing.

Many very smart writers are masters of writing in plain, understandable prose. Paul Bloom comes to mind—he’s an excellent writer, and manages to make complicated research accessible. But despite the accessibility of his writing, I don’t think anyone can claim Paul Bloom is dumb. I look up to his way of using casual, understandable prose as an indication of both his mastery of the material and his empathy for his reader. That’s the kind of good writing I aspire to.

Subscribe now

Dimensionality Reduction

Dimensionality reduction is one of those idiosyncratic concepts that I find useful when thinking about writing. Dimensionality reduction is a method of transforming data so you retain the useful information while reducing the number of dimensions.

That sounds complicated, but we do this kind of thing all the time. Think of political affiliation—while every individual might differ on any number of different issues, we tend to think of a simple 1-dimensional slide from more left-wing to more right-wing. We know we can cut things more finely and understand their positions in more detail if we want, but we also know that a lot of information is carried in that single dimension, and for many purposes that single dimension gives us what we need. Knowing if a politician is left-wing or right-wing allows you to guess their stance on a pretty wide range of issues.

Now imagine I was an alien studying human society and didn’t have the concept of the political left and right, but knew all of the positions on the different issues different politicians and people took. I would be able to look at how opinions tended to cluster and relate in certain ways. If I had enough information about different opinions, and was smart enough with how I analyzed them, I would notice patterns: opinions favoring equality, social change, and government action to fix problems would tend to cluster together (left-wing), and opinions favoring tradition, social order, and individual freedom would tend to cluster together (right-wing).

You need a lot of data before you can crunch things down to this simplified form and accurately capture the patterns, though.

There are a lot of interesting theories and ideas in science and philosophy that are inaccessible because of how much work they are to get at. The language used is hard to parse and the academic version is full of detail that connects it to other findings of interest to academics.

So partially, I see the role of someone like me to act a bit like the digestive system, helping to process all of this information so the gooey nutrients can be absorbed. But doing so requires reading a lot more than what makes it into posts. Sometimes it feels silly to read four different papers for what ends up being a single sentence, but that’s what’s required for dimensionality reduction—you need enough of the higher-dimensional representation first to be able to see the patterns well enough to make sure the lower-dimensional projection is good enough.

One problem is figuring out how much nuance to keep and how much is just adding noise. There’s value in less nuanced takes—you don’t want to miss the forest for the trees. At the same time, it’s easy to end up exaggerating or misrepresenting if you leave out too much. This is a constant tension.

The internet rewards exaggeration. Making things a bit too clear and sensationalized is an easy way to get clicks. It’s easy to imagine getting sucked down a hole and just chasing that engagement. I’m lucky enough to be read by academics—philosophers, psychologists, and neuroscientists—and I care what they think. I’ve consciously tried to court that type of audience and maintain my academic ties. I suspect a lot of writers who “go off the rails” chasing clicks and popularity over integrity just aren’t constrained by their integrity among peers. I want that constraint. I might not get everything right, but most things I write, I have at least one moment of thinking “Hmm, would an expert in X agree with this?”

I try not to let this freeze me up too much. It’s easy to psych oneself out worrying about not saying anything new on the one hand, or saying something an expert would roll their eyes at on the other. But I am grateful to have this constraint, and the gentle nudges I’ve sometimes received from experts when they think I should have put a little more nuance or missing detail in.

Keep me honest.

Subscribe now

Writing Compounds

Forcing yourself to put your thoughts down in concrete terms tests those ideas in a way that merely thinking them doesn’t. It’s pretty frequent that I think everyone’s arguments on a topic are weak, but when I go to put my own thoughts down, I realize my own look similarly fragile.

Relatedly, it isn’t until I start writing that I often notice enormous gaps in my knowledge. Writing is a great way to force myself to learn more and grapple with my own understanding of a topic.

What I didn’t recognize when I started writing is how much this compounds over time. I now have a bunch of topics I’ve written about and have better crystalized thoughts on. It’s not hard, then, to sometimes pull one of those out, if only for a paragraph.

I’ve noticed many of my favorite writers have recurring topics and themes. Cognitive scientist Douglas Hofstadter has self-reference, and science fiction writer Greg Egan’s mind uploads and AI mediated conversations have a similar flavor across many of his stories. When you get a solid grasp on your thoughts on a topic, it becomes a recurring character.

Sometimes this can become an issue—I’ve certainly found books by an author where their second book is a rehash of the first. Hopefully I can avoid that fate (should I ever manage to get an agent and book deal).

Still, being able to “recycle” ideas has come in particularly handy for writing articles for different venues. My articles in Psyche and Skeptic both draw on ideas I first explored in various posts here. It’s much easier to write those sorts of things after having worked through the idea, allowing me to put these disparate already-worked ideas together in a different combination. Instead of smelting all the raw ore, I’m starting with refined materials.

This means I’m able to revisit topics more than once. I find this a great way to combat perfectionism—I know any time I write about a topic, I can always come back to it later, tackling it from a different angle or adding new ideas to it. I don’t need to exhaustively cover it the first time. Even though each post goes through multiple rounds of editing before publishing, everything is a rough draft.

Including this.

If you enjoyed this, please hit the “Like” ❤️ button, restack, or share this article to help others find it.

If you enjoy Cognitive Wonderland, consider supporting it by becoming a paid subscriber at whatever level feels comfortable for you.

If you’re a Substack writer and have been enjoying Cognitive Wonderland, consider adding it to your recommendations. I really appreciate the support.

Subscribe now

Share

Leave a comment

From the back cover: Most of us are aware of the hive mind—the power of bees as an amazing collective. But do we know how uniquely intelligent bees are as individuals? In The Mind of a Bee, Lars Chittka draws from decades of research, including his own pioneering work, to argue that be…

Read more

While it’s easy to beat yourself up about not having enough willpower to simply make the change, the evidence shows behavioral change is hard for pretty much everyone. Most smokers want to quit but very few succeed, most dieters regain whatever weight they lost, and anyone who has been to the gym in January knows how quickly New Year’s fitness resolutions are abandoned.

Behavioral change feels so trivial, and we often see people who don’t struggle with the same behaviors we do, so it feels easy to believe if we just had more willpower or had the right trick, we could get ourselves under control. This understandable impulse is why there’s a cottage industry of TED Talks by people promising easy solutions that will change your life. Self-help is a multi-billion dollar industry.

A big problem with the easy solutions, though, is just that: they’re easy. The idea that we just need to form the right intentions or take the right mindset feels right, and might work for a little while, but as easy as these internal changes are to make, they’re just as easy to unmake, and leave us right where we started.

One common bit of advice is to take a “growth mindset” approach to doing difficult things. The issue is, after multiple large meta-analyses, there’s not much evidence growth mindsets actually do anything. It’s worth noting studies investigating growth mindset interventions involve actually sitting subjects down and having them do an exercise like writing an essay, not just watching a Ted Talk or reading an article. This is consistent with the broader findings that behavioral interventions that target beliefs and attitudes have very small impacts on behavior.

The idea that we just need more willpower or a simple mentality shift ignores the actual determinants of behavior.

Subscribe now

Context is King

In the early 1970s, there was a scare about the rate of opiate addiction among servicemen in Vietnam. The United States was about to withdraw 200,000 troops, and there were reports that many of them were addicted to heroin. There were media portrayals of the “Addicted Army”, and worries about what such a huge addicted population returning would mean for America.

These worries weren’t overblown—it turned out about 43% (!) of veterans admitted to using some form of opiate in the year prior to returning home. 20% of veterans were heavy enough users to experience withdrawal symptoms.

What happened when they returned home? Basically all of them just stopped using. In the year following returning home, only 1% became re-addicted.

The low rates of relapse among Vietnam veterans were initially met with disbelief, but they held up under scrutiny. Why the extreme change?

Researchers interviewed the veterans to find out what made them turn to drugs in the first place, and why they didn’t continue using when they returned home. The reasons for using them while in Vietnam were what you might guess: boredom, homesickness, and being unable to sleep. On top of that, it was cheap, others were using it, and it was easily accessible.

But on returning home, they said they didn’t re-engage with it because they feared being arrested and knew it was strongly disapproved of by friends and family.

The returning vets didn’t just go home and not receive any help—those that tested positive for opium in their urine were given a 2-week detox period before they were allowed back.

But the relapse rate among inpatient centers treating people with addictions is extremely high—one study estimated the relapse rate of opiate users at 91%. This isn’t an apples-to-apples comparison with the Vietnam veterans (the form and frequency of use among the populations likely differed, for one thing), but it highlights how hard this change is. Removing the physiological cravings isn’t enough to prevent falling back into old patterns when you’re returned to the same context you initially started using an addictive substance in.

The difference between people treated at centers for their addictions and the returning soldiers is the complete shift in context for the soldiers. The incentives (social approval/disapproval) changed, the problems they were trying to solve (boredom) changed, and so did the friction (cost and accessibility) of acquiring drugs.

The Path of Least Resistance

For the returning veterans, opium wasn’t as easily accessible back home. If they sought it out they needed to find a new source, and it was more expensive. There were a bunch of new barriers to jump through, so only those very motivated to get drugs would jump through them.

Compare this to people coming home from a treatment clinic—they already know how to get the drug. There may be no change in how difficult it is to obtain opiates when they return.

Our environment sets which behaviors are easier or harder. Part of the difficulty with many of the behaviors we would like to be doing more of—eating healthier, cheaper meals made at home instead of eating out, or starting a gym routine—is that the alternative is just too easy. It’s always easier to just not go to the gym.

We often take the path of least resistance—especially when we’re stressed or busy. But one of the nice things about this is that we can change the relative friction of different paths.

Those that have fruits and vegetables around the house tend to eat fruits and vegetables at meals. This is obvious, but what you buy at the store during your weekly grocery trip affects what is easy and accessible all week. Similarly, if you prep lunches for the full week, you’ve changed the relative friction between eating a home-made lunch versus eating out.

The same principles hold for other behaviors we want to change. Many of us use our phones more often than we would like. People claim it’s an addiction, but the real problem is the lack of friction. Phones provide entertainment and are incredibly frictionless. They’re with you at all times, you can simply whip it out and start scrolling any time you’re the slightest bit bored. Turning on the TV to watch Netflix is a gargantuan task compared to phone use.

It’s no surprise, then, that studies show increasing the friction to using your phone leads to lower phone use. Apps that introduce a brief delay before an app opens reduce the time spent in those apps. Hard lockout apps don’t seem to work as well as ones that add friction to actions taken, for the simple reason that usually they have some frictionless way of dismissing them.

Anecdotally, people talk about scrolling their phones when they wake up in the morning and fixing that by just leaving their phones somewhere they can’t reach from bed. No one is going to get out of bed, get their phone, and get back in bed. When you add friction to the phone-in-bed habit, you reduce the behavior.

More broadly, research on the efficacy of various policy-level interventions meant to promote public health show that policies that promote access (i.e. reduce friction) have the largest effect, while policies that target knowledge or beliefs have limited to no impact.

This isn’t to say that our behavior is all determined by whatever is easiest. But especially with long-term, ongoing behaviors, you have to consider what you default to when stressed or not thinking about it. Which leads us to another major determinant of our behavior: habits.

Subscribe now

Habitual Creatures

It’s a cliche to say that we’re creatures of habit, but… we’re creatures of habit.

Habits are associations between a context and a response. Given a specific context we do a specific habit without much thought. When I wake up, I grab some clothes from my dresser, head downstairs, make tea, and then come to my office to do some writing. Typically I don’t consciously deliberate about any of these things.

When context is stable, our past behavior becomes very predictive of our future behavior. A time-use study found that 43% of the time, college students were doing something they do everyday and usually in the same location. Many of these behaviors were things like hygiene, getting ready for the day, getting ready for the night. Most of the time while doing these behaviors, the students were thinking about something else. This leads to one of the benefits of habits: when you automate behavior, you don’t think about them, so you free your mind up for other things.

Habits are a double-edged sword. Because they’re reflexive and based on context, without a change in your context they push you toward habitual behavior. We tend to fall back on habits in times of stress—when we’re preoccupied, we rely on the automated responses we have.

If the habits we have are aligned well with what we want, they can keep us doing those actions without thinking about it. If you go to the gym three times a week, and have an established routine around it, you’re more likely to keep doing it even if work is suddenly stressful. But if our habits are in conflict with what we want, they can work against us—if you’re trying to quit smoking, but have a habit of taking a smoke break at a particular time, when life gets stressful you’re more likely to fall back on that habit. Your mind is too preoccupied to go against the flow of the automated routine.

Habits don’t rule all of our behavior. Our conscious intentions are important determinants of many behaviors. Making an intention more specific, like writing down a plan, makes people more likely to do things like get vaccinated or vote.

But the hard thing about many of the behaviors we want to change is that they are ongoing. Research has shown that intentions are strong predictors of one-off or occasional actions, but not actions that are ongoing. Motivation and intentions are fleeting, and rarely keep us doing something long enough to actually drive lasting change in our habits.

Easy tricks that change your intentions or attitude can help for the short term but not the long-term. Reviews of interventions designed to promote healthy behaviors all seem to have the same message: it’s pretty easy to get people to exercise regularly or eat healthier during the intervention, but when you check back in later it’s rare these new habits “stuck”.

A recent megastudy targeted 61,293 members and used 30 scientists to design 54 different four-week programs or interventions to encourage exercise. While almost half of these interventions led to small increases in gym visits during the 4-week intervention period, this shrunk to almost nothing looking at the 10 weeks that followed the intervention.

Long-term weight loss is difficult. Study after study shows a specific trajectory of calorie-restriction interventions: an initial loss, followed by a gradual regaining of weight. The reason is pretty simple: appetite increases during weight loss. You have all of the normal environmental factors that lead to obesity in the first place—our environments tend to be low activity, high calorie food is abundant. Researchers have suggested that with the increased appetite, there may be a slow, unconscious drift up in consumed portion sizes. They go on to say:

Long term behavioral changes and obesity management require ongoing attention.

[...]

The importance of long-term intervention has been codified in the obesity treatment guidelines, which state that weight loss interventions should include long term comprehensive weight loss maintenance programs that continue for at least 1 year.

Despite what you might hear in a TED Talk or in a self-help book, there isn’t some simple way of changing your mentality to suddenly approach everything differently. If there was, behavioral scientists would have found it and the failure rate of obesity interventions wouldn’t be so high.

There’s an idea that some people just need more self-control. If they could just overcome those temptations, they would be better off, and perhaps they can develop self-control like a muscle.

Unfortunately that’s not how things work. It isn’t the in-the-moment self-control that drives successful outcomes, but “trait self-control”—which turns out to be the same thing as conscientiousness. Conscientiousness is a personality trait characterized by organization, dutifulness, and goal-directed behavior. Those with high conscientiousness actually engage in less moment-to-moment self-control. It isn’t that they’re better at resisting temptation, they’re better at avoiding it altogether. And unfortunately, we don’t know any good way of changing conscientiousness for the long-term, so if you’re hoping to become more conscientious so you can change your behaviors long-term, behavioral science can’t do much.

Our environmental context creates our incentives, the friction to doing things, and contains the cues that drive our habits and routines. This might all sound pessimistic, as if we’re all helplessly driven by our environment. But we have a lot of control over our environment. Our motivation is probably better spent shaping the environment, because we are finite creatures with finite attention. When motivation inevitably fades, we’ll be left with whatever context we’ve shaped.

Subscribe now

Change and Fresh Starts

It’s possible to change your behavior through internal changes, but they’re not easy. Cognitive behavioral therapy involves working with an experienced therapist over a period of time to reshape our habits of behavior or thinking. It takes multiple sessions and is personalized. Inner changes are hard and require time, support, and sustained effort.

As leading habit researcher Wendy Wood puts it:

[C]hanges in intentions, beliefs, and attitudes have little long-term impact on habitual responses. Enduring changes require directly influencing behavior performance through intensive therapeutic interventions or structural changes of new reward systems, context shifts, and friction on unwanted responses.

Changing the reward systems, context cues, and friction in our environments can be hard—if it were easy, behavioral change would be easy and everyone would do it (and there would be no more money for those hawking easy “fixes”). For the addicted Vietnam veterans, all of these things changed for them simply by going back home. The rest of us are unlikely to have a similarly disruptive event happen, but we can create or take advantage of smaller scale opportunities in our own lives.

Big life changes, like getting married, starting a new job, or moving, all lead to new routines and contexts. These changes open up an opportunity where habit formation is more directly related to our intentions. Successful attempts to change behavior more often involve moving to a new home than unsuccessful ones. So it seems possible to take advantage of these transitions to set up routines and habits better aligned with our goals.

Moving (or getting married) just for the sake of shaking your life up enough to establish new habits probably isn’t a great solution. But we also mentally think of certain temporal landmarks as opportunities for “fresh starts”. It’s been suggested that events like New Year’s lead to a motivational boost (and aspirational “New Year’s resolutions”) because they give a landmark to relegate past imperfections to. This mental accounting might give an opportunity to change our context.

Of course, this is an easy “inner change”, hence the quick decline in gym attendance past January. But these boosts in motivation, wherever they come from, can still help, since attitude shifts can lead to new habit formation. They need to lead to repeating a behavior enough in a stable enough context for it to become the routine. You can’t rely on motivation alone to perform a behavior indefinitely, but you can use it to change your context to improve your chances of doing it enough for it to stick.

The goal is to make changes to your environment that are costly to undo. If you automate savings, cancel a subscription, or remove an app on your phone, you now have to actively do something to set the undesired behavior as your “default”. Eliminate friction from things you want to do, add friction to things you don’t want to do, and make those changes costly (in effort, time, or money) to reverse.

Anything that you make costly to reverse is going to pull you in the right direction. Contrast this with easy fixes like “taking the right mentality” or “forming an intention”: these are easy to reverse. They offer no friction against changing your behavior back, and that backsliding is sure to happen as soon as you need your attention on something else.

All of this is easier said than done. If there was an easy way to change context to encourage healthier habits, it would be known by now and we wouldn’t continue to see such weak effects in studies of behavioral interventions. But what’s more important is to ignore the stuff that doesn’t work—no single neat trick is going to change your behavior. Sheer motivation or willpower likely won’t get you there. Changing your context—so you don’t face the same habits and frictions that shaped your behavior in the first place—might.

If you enjoyed this, please hit the “Like” ❤️ button, restack, or share this article to help others find it.

If you enjoy Cognitive Wonderland, consider supporting it by becoming a paid subscriber at whatever level feels comfortable for you.

If you’re a Substack writer and have been enjoying Cognitive Wonderland, consider adding it to your recommendations. I really appreciate the support.

Subscribe now

Share

Leave a comment

In a previous article, I talked about how interactions between ants create a closed-loop that regulates how many go out to forage. With nest building, there’s an important difference: many of the signals ants and termites use to tell them what to do don’t come from direct interactions with their nestmates. Instead they come from the environment that the insects are collectively interacting with.

For example, in a study of the nest building of the ant Lasius niger (black garden ant), researchers investigated how the ants coordinate. They found the ants start by picking up soil and forming it into pellets for building. Then, if there is a nearby deposit of pellets, they add it to the pile; otherwise they just put it down wherever (which initiates a new pile). This very simple rule ensures the pillars are spaced out properly and allows ants to cooperate while building up pillars without a central plan.

When the pillars are tall enough (judged, adorably, by whether they are taller than the ant is tall), the ants start extending the top of the pillar outward to create a roof.

What I want to explore here is how the ants use the environment itself to coordinate. This concept has deep implications, but I hadn’t even heard the name for it until very recently: stigmergy.

Subscribe now

Stigmergy

Instead of having a central plan or direct communication, the ants building their nest coordinate through altering the environment. One ant drops a pellet in a spot, and the pellet acts as a signal for where to add more pellets.

This is one form of stigmergy:

Stigmergy is an indirect, mediated mechanism of coordination between actions, in which the trace of an action left on a medium stimulates the performance of a subsequent action.

In the ant pillar case, the trace is the pellets and their location, and the medium is the nest site itself.

The subsequent actions are based on the environment—you can imagine the ants following a sort of conditional statement:

If there is a pellet pile nearby, then add pellets to the pile (until it’s higher than the length of your body)

While stigmergy can allow many ants to coordinate together, it can also allow an agent to coordinate with itself over time.

Self Stigmergy

Maybe the simplest example of stigmergy is a thermostat. Based on the current temperature, the thermostat turns on a heater or air conditioner. Doing so changes the temperature, leaving a “mark” on the air.

This setup provides a bunch of benefits: the thermostat doesn’t need any kind of complex inner model of how its actions affect the temperature. By just constantly reading in the current state of the environment, it keeps in perfect sync with what it’s meant to affect.

We find single-agent stigmergy in nature, as well. Enter our old friend, the slime mold.

I’ve previously talked about how slime molds, which are giant single-celled organisms, are able to use local rules to perform smart decision-making. Their veins explore the environment looking for food. Veins that find food are reinforced, becoming thicker. All veins are in competition for cytoplasm with all other veins, so those that don’t find food gradually diminish.

With these simple feedback rules, slime molds can find efficient routes through mazes and make various trade-off decisions, despite lacking a brain.

But here’s the wild thing: slime molds also have a form of memory.

If a slime mold is trying to forage efficiently through an environment, it’s going to do a lot of exploring—reaching out its tendrils searching for food sources. But without memory, it could end up searching an area it’s already been in, wasting precious resources.

Luckily slime molds have a solution: they leave behind a slime trail, and use the presence of slime as a cue that they should explore elsewhere.

So instead of needing to build up some mental map to keep track of where it’s been, the slime just uses the world as its own map. By leaving a mark in the environment, it changes its own behavior in that environment, allowing for more efficient foraging.

We humans with our complicated brains tend to keep our memories inside our skulls, but not always.

Subscribe now

My Stigmergic Life

As a kid, I remember watching this Sesame Street bit where Ernie used strings tied around his finger to remind him of something—the first string was to remind him he had a string tied around his second finger, the string around the second finger reminded him he had a string tied around his third finger, and so on, until the last string was supposed to remind him they were out of string. Ba dum tss.

The idea of tying a string around your finger as a reminder stuck with me. I started putting odd things on my bedside table to remind myself I meant to do something as soon as I woke up—like a pencil on top of my alarm clock’s off button to remind me I had to stick my assignment in my backpack.

I was making a mark that would spark a rule. If something weird was in my environment, then try to remember why it was there. If I remembered why it was there, I remembered what I had to do.

As a parent, I’m often more slime mold than human. One of my jobs in the morning is to get the kids’ bags all packed and ready for preschool. There’s a bunch of steps to this—fill their water bottles and put them in their backpacks, put milk in their thermoses and put them in their lunch bags, put utensils in their lunch bags, and so on. But I do this while being interrupted every 30 seconds by other demands from the kids, like them needing me to cut a peanut butter sandwich into the shape of the corndog king.

Relying solely on memory, I wouldn’t be able to keep track of where I am in the process. Instead, I just rely on the environment, and take simple steps based on what I find when I look at the kitchen counter. A filled water bottle? Great, next step is to put it in the backpack. Each step I take leaves a mark on the medium of “stuff on my kitchen counter”, tracking where I am in the process.

The downside of relying on stigmergy is that it’s sensitive to disruptions in the environment. Washing away the slime of a slime mold would wipe away its memory. The Eumenid wasp, which constructs a sort of mud funnel as part of its nest, will restart from step one and build a second mud funnel if a hole is created at the right time in the nest building process—the hole acts as a cue that it’s at step 1.

Similarly, if the lunch bags got zipped up before I put utensils in them, I would be missing the environmental cue to put the utensils in, and my 5-year-old would give me an earful when he got home.

Subscribe now

The Retracted Mind

We all use stigmergy in various forms constantly. We leave out a form on the kitchen table to make sure we remember to fill it out. We use a digital or physical calendar to prompt us when it’s time to do something.

There’s common writing advice to “just start writing”. The idea is you can edit a bad draft, but you can’t edit a blank page. This works for a lot of people—and is a form of stigmergy. The draft is more than just a worse version of the final product, it’s also a signal that kicks off a different kind of process—a set of responses to the contents of the draft that wouldn’t be possible without those marks on the medium.

If you’re familiar with the idea of the extended mind, this might all sound familiar. It’s the idea that some parts of our environment can be so fundamental to cognition, we should consider them part of the mind. Here’s a summary of the key thought experiment:

The fictional characters Otto and Inga are both travelling to a museum simultaneously. Otto has Alzheimer’s disease, and has written all of his directions down in a notebook to serve the function of his memory. Inga is able to recall the internal directions within her memory. The argument is that the only difference existing in these two cases is that Inga’s memory is being internally processed by the brain, while Otto’s memory is being served by the notebook. In other words, Otto’s mind has been extended to include the notebook as the source of his memory. The notebook qualifies as such because it is constantly and immediately accessible to Otto, and it is automatically endorsed by him.

The usual reading of this thought experiment extends the mind outward, so Otto’s notebook becomes part of the mind. But we can invert this reasoning instead: maybe we should consider Inga’s use of memory as a form of stigmergy.

At any given moment, the cognitive processes operating now are dependent on memory traces from past processes. Those past processes have no direct contact with the current ones, but they make marks in a shared medium—memories stored in the synaptic connections of the brain. The current process “picks up” those memory traces and does something different because of their presence.

Cognitive scientists often implicitly invoke stigmergy in the analogies they use. In Global Workspace Theory, they talk of a “blackboard” that many different cognitive processes read and write to. Working memory is described as having a “visuospatial scratchpad“. Engrams, the physical alterations in the brain that encode a specific memory, are the marks made in the medium of the brain, altering the later actions taken by processes.

The phonological loop is a part of working memory that stores speech that we hear for short periods (1.5-2 seconds). If you’ve ever found it easier to count up a bunch of items aloud rather than in your head, it’s because you’re relying on the phonological loop to hold the previous number. You’ve made a mark in the phonological loop to hold your place, and you know, for each number, what comes next. If seven is in my phonological loop, then grab the next item and say eight. How different is this from making a tally on a page, or using one’s fingers to keep track?

Our minds are full of different parts constantly making marks on each other, and not just for memory. From the perspective of any one process, the rest of the brain is a medium to mark up. The brain is constantly changing—connections between different neurons strengthening or weakening through various biochemical processes, allowing us to learn. These are all marks being made in the medium, memory is just the most obvious case.

Above, I mentioned how stigmergy can fail—if a mark in the medium is disrupted (like a hole added to a wasp’s mud funnel or the lunch bags getting zipped up), it can lead to a misfiring of the process relying on those marks. But our own internal stigmergy takes place in an often easily disrupted medium. When my kids desperately need me to answer a question about whether sea snails are real, it wipes out what was in my working memory. That medium just isn’t robust for when kids are around. So I instead rely on the more reliable medium in that situation: the kitchen counter.

If you enjoyed this, please hit the “Like” ❤️ button, restack, or share this article to help others find it.

If you enjoy Cognitive Wonderland, consider supporting it by becoming a paid subscriber at whatever level feels comfortable for you.

If you’re a Substack writer and have been enjoying Cognitive Wonderland, consider adding it to your recommendations. I really appreciate the support.

Leave a comment

Share

Subscribe now

If you do not want to receive posts like this about Book Club logistics (voting for books and learning what was selected), you can turn it off in your account settings:

Also, if you haven’t already, make sure to check out the discussion thread for our April read, The Mind’s I.

Now on to the voting. Her…

Read more

I was shocked by the credulity of the reporting. A cursory read of the scientific papers or asking the researchers for any details would have seriously deflated the story. And yet the original viral news story contained none of these details, which led some readers to conclude they must have somehow hooked up the brain cells up to a camera and controller.

This isn’t what was done. A reinforcement learning algorithm was given direct information about the gamestate and played Doom by inputting their commands through a brain cell network. The brain cells acted a bit like a controller for the reinforcement learning algorithm. If they were helping instead of hindering the reinforcement learning algorithm play at all, they likely weren’t doing much—the brain cells on a chip struggle to learn simple mappings.

So what the hell was the science journalist who wrote the article doing? I understand giving the article a click-bait headline and then giving the important details in the article—at least then people who are engaged enough to read the article learn the truth. But what’s the point of a 600-word article that misleads more than it informs? Is it just a paycheck for the author? Don’t they take pride in helping people understand the science, or are they just doing whatever maximizes clicks?

The pattern of overhyped popular science articles can’t be blamed entirely on journalists. University press releases, which generally form the basis of the news stories, frequently exaggerate the research. One study found about 40% of press releases contain exaggerations. When the press release contains exaggerations, most press articles contain exaggerations. When the press release does not, most press articles do not.

Press releases aren’t the only issue, either. Often the exaggerations exist in the abstracts of papers as well.

So it isn’t just journalists doing the hyping, but scientists as well.

That’s certainly the case with the Doom project. It was put together by a private company, so obviously they have incentives to hype the project. But the video they released has no mention of the reinforcement learning algorithm that actually does the heavy lifting. Over on the Doom Neuron community they set up on Twitter/X, they actively suppress any pushback correcting the hype.

If I may indulge my cynicism here for a moment, this all makes sense. Researchers in general and those at a private company in particular have incentives to exaggerate the impact of their work. You’re more likely to get funding and attention if people think your work is important. Similarly, science publications need to draw attention to get readers, so surely they want their writers producing work that gets clicks.

The incentives are bad, but the picture isn’t totally bleak. For one thing, scientists are one of the most trusted groups in society—this is reassuring but also feeds into the hype issue.

Public Trust in Scientists

I often hear the sentiment that scientists need to change their act because “just look at the state of public trust in science!”

As much as I think there’s lots of room for improvement in how scientists interact with the public and how science gets disseminated to the public, it’s worth taking a realistic look at how things stand. For one thing, it is hard to find another group in the United States that is more trusted by the public—only the military is trusted more to act in the best interests of the public.

For another, to the extent there is distrust in scientists, there’s really two major factors: 1) political partisanship, and 2) the Covid-19 pandemic.

This figure from Pew nicely sums it up:

So two somewhat unsurprising things: 1) confidence in scientists is lower among republicans than among democrats, and 2) trust has been lower since the COVID-19 pandemic, where public health scientists had to make a bunch of calls in a state of uncertainty.

I could say a lot about this, but for the purposes here, the point is the public generally trusts scientists, and the areas that science communicators need to be careful about are where science intersects with cultural identity and large-scale public issues. This raises important issues for scientists trying to tackle difficult polarized topics like climate change or vaccines, and there’s a lot to be said about that. But my concern here is different: While it’s good that scientists are trusted, it produces a secondary problem when this combines with the above incentive structure to hype things up.

Subscribe now

The Hype Machine

Scientists, university press releases, and articles appearing in popular media often exaggerate findings. There are structural incentives for scientists to overstate the impact of their research, and for journalists to provide sensationalized versions of stories.

The high trust in scientists also makes it easier for various credentialed “science influencers” to exploit that perceived authority to spread sensationalized hype or blatant misinformation online.

To some extent this is self-undermining. Scholars have argued this overhyping erodes public trust, even when the hype isn’t believed. Suspecting a science communicator is exaggerating can lower your trust in science communicators.

If you scroll through social media, you’ll be exposed to a ton of science claims, ranging from click-baity headlines to influencers making unsupported wellness claims. I suspect if you took a random sampling the bulk of them would be exaggerations and straight up falsehoods.

However, the problem isn’t just one of individual actors and incentives. There’s a mismatch between the desire for news and the way science works.

Millions of peer-reviewed articles appear annually. Individual studies are rarely important. Every study has flaws, because every experimental method has flaws. Every measurement instrument is a flawed instrument. Every sample is a flawed sample. Every researcher is a flawed researcher. This isn’t to criticize science or researchers, it’s just to point out that, the vast majority of the time, a single study doesn’t prove anything conclusively. Science is a gradual process, with multiple different studies coming from different angles to cover the weaknesses of previous studies and converge on a result.

The unfortunate dynamic is that popular science writing tends to focus on individual studies—it isn’t “news” unless it’s “new”. Stories written about new studies tend to leave out important limitations of the studies, likely because that detracts from the story rather than making it more interesting.

In my experience, it’s much more common to run into science stories about single studies rather than about reviews. Systematic reviews attempt to pull together a bunch of different related research on a topic and give a high-level view of what is known and what is yet unknown. This is fertile stuff for science writing, begging to be picked up and turned into a story interesting enough for a curious audience. I don’t see much of that happening, though.

Subscribe now

The Role of the Debunkers

So we have this funny situation: scientists are trusted, science is mass produced, scientists are incentivized to hype their work, and science journalism is incentivized to at least go with the hype. The result is a deluge of science content, much of it of dubious quality.

But there’s good news: debunking works.

Specifically, the approach of stating the misconception, explaining what’s wrong with it, and providing the correct information has been shown to work across a wide variety of topics. Simply put, confronting misunderstandings head on seems to work. This is also true of neuromyths specifically.

But this leaves out what I think is another essential function of debunking: holding the journalists and scientists who overhype accountable.

I think it’s good and healthy for there to be an ecosystem of accountability. When there are so many incentives pushing individuals to exaggerate, one thing that can moderate that impulse is a strong group of peers who judge and call out those claims that go over the line. If a scientist expects to lose respect by publishing some fanciful exaggeration about their research, they’ll think twice about it. Journalists who write misleading popular press articles should also feel some judgment.

I see this as just part of the ecosystem that holds exaggerations in check. It not only keeps engaged members of the public better informed, but helps retain trust in science by acting as a constraint counterbalancing the incentives to hype.

Knowing you might be called out for saying something ridiculous is good. I’m happy to be a small part of the accountability mechanism of the science communication world. I’m also happy to go out of my way to form connections with professors and other experts in the things I write about—knowing I have various professors of neuroscience and philosophy reading my writing makes me a bit more humble about what I say. I’ve benefited when an expert has pushed back a bit to add some nuance to something I’ve said.

That said, not all science writing is, or should be, debunking.

Subscribe now

Science Writing More Generally

Stefan Van der Stigchel is an attention researcher who decided he wanted to write a book for a popular audience. He believed his scientific knowledge could help people who wanted to improve their concentration, and there was a lot that he knew that the public didn’t. He also saw a lot of self-help books out there that were inaccurate and wanted to help people find more reliable information.

So he wrote a book. In an academic article he wrote about the experience, he talked about the twin difficulties of writing for a general audience. On the one hand, his audience is expecting definitive answers and clarity. On the other hand, he worried about his colleagues’ judgments over leaving out too much detail. In an area where the science is often provisional and it’s unclear how it translates to real life, he had to make decisions on what nuances to leave out and how to make his message clear without losing accuracy.

If you read the reviews of his self-help book, people weren’t happy. It has a low score on Goodreads (3.3/5), and here are some representative quotes from a few reader reviews there:

I was hoping for techniques and information on improving your concentration, instead it begins with a load of studies on memory.

[I]f you were hoping for techniques to build the muscles of focus, you’ll be profoundly disappointed.

[I] was looking for direction/action items on improving my own concentration and this book didn’t really have that.

There’s a real tension here between sticking to the facts and writing something satisfying for a general audience. As Stefan says, he was writing as a response to self-help books he thought weren’t science-based. He was attempting to put a better alternative out there. But by sticking closer to the facts, his book was less attractive than the less factual competitors that were willing to be more definitive. If you just want a book that will tell you the definitive one simple trick to make you concentrate better, you’re not going to want a book any reputable scientist would write.

There’s a real tension: science writers want to stick to the facts, but they are competing with books and articles that have no such constraint.

Scientists that take on the hard job of writing honestly for a general audience deserve more credit than they get. If they don’t try, the space is just ceded to whoever manages to capture the public’s attention, with or without scientific accuracy.

Why I Write

I first got into reading science and philosophy when I was in high school. I would skip class to go to the bookstore, and it was there that I first encountered books like Stephen Law’s Philosophy Gym and Charles Darwin’s On the Origin of Species. These books filled me with wonder. They were idea books, giving me new concepts and perspectives to understand the world. They allowed me to look at things in the world that I took for granted in a new way.

Hyped up science stories that distort people’s understanding of the world undermine this wonder. Telling people “brain cells on a chip can play Doom!” might generate a quick “Oh wow!”, but then you move on. Even if it were true, it doesn’t lead to a deeper understanding. It’s a “neat fact” that doesn’t help someone understand the how or why. It doesn’t extend their island of knowledge.

When I think about who I write for, it’s for that teenager who skipped school to find wonder in the bookstore. The person seeking ideas and new ways of seeing the world and making an honest attempt at finding the wonder of legitimate understanding.

If you enjoyed this, please hit the “Like” ❤️ button, restack, or share this article to help others find it.

If you enjoy Cognitive Wonderland, consider supporting it by becoming a paid subscriber at whatever level feels comfortable for you.

If you’re a Substack writer and have been enjoying Cognitive Wonderland, consider adding it to your recommendations. I really appreciate the support.

And yet, scientific psychology is full of findings, established through careful investigations, that are surprising.

Paul Bloom’s Psych is a tour through what we’ve learned from scientific psychology, drawing on the introductory course Bloom taught at Yale.

It’s easy to think of memory as a sort of video recording that gets stored and then retrieved later. But one of the most robust findings in psychology is that memory is constructive. If you show people a string of words like “bed, rest, awake, tired, dream, wake, snooze, blanket, doze, slumber, snore, nap, peace, yawn, drowsy”, and then ask them to recall the words, many will confidently claim “sleep” was presented. It fits so well with the vibe it seems like it should be on the list.

This doesn’t just apply to remembering arbitrary things. The famous psychologist Jean Piaget had a vivid memory of being nearly kidnapped as a toddler and being saved by his nurse who fended off the attacker. He could even visualize the scratches on the nurse’s face from the fight. Only many years later did he learn the story was made up by his nurse, who confessed she’d invented it to get a reward from his parents.

Subscribe now

Small changes in how questions are asked can change how we remember them. After watching a video of a car hitting a pedestrian, asking “Did you see the broken headlight?” as opposed to “Did you see a broken headlight?” leads people to remember seeing a broken headlight. Findings like these have led to a revolution in the law, where it’s now understood that police interrogations can change or create memories.

It’s easy to have the same naive attitude towards perception. Perception isn’t a passive process where the world is transparently conveyed to us. Our expectations affect what we perceive.

When we read, the bottom-up information of the letters we see obviously informs what we see. But so do our top-down expectations. In the above image, you probably didn’t notice the extra “the”.

Similarly, the top-down expectations from context can change how we recognize letters. Consider this example:

Presumably you have no problem reading “THE CAT”. But the ‘H’ and the ‘A’ are the exact same symbol. The top-down influence of the context of the word influences how we perceive the individual elements.

Psychology hasn’t just taught us interesting things about memory and perception. It also gives us insight into what is innate and how minds develop.

As a parent, I’ve watched my kids go from infants who just scream and randomly flail their limbs to little people capable of conversations and making surprising inferences about the world. How do kids make this transition?

We’ve learned that children are a lot more sophisticated than we might have thought. They pick up on statistical regularities in language, slowly learning to tease apart words and understand speech. They have an understanding of how physical objects act, and show surprise when objects don’t act in the way you would expect—for example, if one object is hidden behind a screen, but two objects are revealed when the screen is lifted, they look for longer, showing they’ve noticed the discrepancy. Contrary to the idea that they lack object permanence, babies have a decent understanding of how physical objects work.

Despite all of this, the scientific study of psychology requires some humility. There are so many things that we want to understand that are difficult to tease apart.

What makes people happy? This might sound like a straightforward question, but when you try to study it, there are constant struggles with interpretation. Volunteering correlates with happiness, but does volunteering make us happy or does being happy make us more likely to volunteer? Or does being a kind person make one both happier and more likely to volunteer? These sorts of causal questions are difficult to disentangle.

Harder still are questions about heredity. We know a bit about the degree to which different traits are caused by genetic factors, or by a shared environment (like family). But a huge portion comes from “non-shared environments”, which could be serendipitous encounters unique to the individual that aren’t shared with siblings. For the portion of traits that are affected by genes, we have a very limited understanding of how genes—which code for proteins—can affect large-scale things like behaviors and personality traits.

Our handle on mental illness—whether our current labels really carve out the natural kinds or lump together very different underlying mechanisms—is poor. Our treatment options are similarly impoverished. Mental health treatment helps many people, but it’s easy to imagine a future where we have a deeper understanding of these conditions and many more people could be helped effectively.

Our grasp on how the mind works has unveiled a lot, but we should be humble that there remain so many unknowns.

Subscribe now

The Replication Crisis

An additional reason for humility is the replication crisis.

Researchers tried to reproduce many of the classic findings in psychology and discovered many of them didn’t replicate. These include big, landmark psychology papers that had been taught in introductory Psychology classes for years.

Many people have taken the replication crisis to mean that we can’t trust psychology at all. This is wrong. There are plenty of well established results in psychology. One way to read Psych is as a survey of what survives post-replication crisis scrutiny. Bloom explicitly addresses the replication crisis and some of the more tangled results in, for example, the priming literature.

Priming is when exposing someone to one thing influences how they respond to something else. This is well established in some domains—if I show you the word “doctor” you’ll be quicker to recognize and respond to the word “nurse” afterwards. The concepts are linked, so calling one to mind makes the other easier to come to mind.

What’s more controversial is behavioral priming. One famous study claimed that exposing people to words about old age (like the word “elderly”) led them to walking more slowly down a hallway. This result has not held up.

Other priming studies claim that sitting on a wobbly workstation makes people think their relationships are less stable, holding a resume on a heavy clipboard makes you think better of the applicant, holding a cold object makes you feel more lonely, and so on.

The replication crisis has not been kind to social priming. To the extent any of the research holds up, the effects are small and unlikely to play any significant role in our everyday behavior. For better or worse, we can’t blame much of our behavior on subtle environmental factors.

How did some psychology research go so far off the rails? Some of it is due to outright fraud—Bloom talks about Diederik Stapel, a researcher who fabricated a huge number of studies. When he talks about his motivations, Stapel describes being bored of giving talks in the small-to-medium rooms at conferences. He wanted to move beyond that. Seeking fame and glory led him to simply make things up.

But the vast majority of bad studies aren’t due to outright fraud. As I’ve talked about before, it’s mostly questionable research practices. Statistics only work when you use them with a specifically defined question. But there’s a few different ways you can look at the data, so researchers often end up asking a bunch of different questions of the same data—and only report the questions that give interesting results.

Imagine you’re trying to find out if holding something cold makes people feel more lonely. You run the experiment with a few different cold objects and a few different people. You run the statistics. No greater loneliness in those that held the cold thing than the not-cold thing. But maybe you remember there was that one guy who you’re pretty sure wasn’t paying attention. You try excluding his data. Still not a significant result. Then you remember a couple of the cold objects weren’t as cold, so you try excluding that data. You try that with and without the “not paying attention” guy. Then you decide maybe it only works on women, so you try separating by gender. Male or female, with or without the less-cold objects, with or without the participants not paying attention. Before you know it, there’s a combinatorial explosion of different ways you could analyze the data. Each of them feels defensible.

Statistics don’t tell you if something is definitely the case. Statistics tell you whether there’s a probabilistic difference from what you would expect by chance. If you keep asking slightly different questions, you’re bound to eventually find one that, by chance, is “statistically significant”. But it’s not likely to replicate, because it’s just a fluke in the data that you found. Cast a wide enough net for flukes and you’ll find one.

I think people generally knew these kinds of questionable practices were bad. As one set of critics that Bloom quotes put it: “Everyone knew it was wrong, but they thought it was wrong the way it is wrong to jaywalk.” But, the critics say, “it was wrong the way it is wrong to rob a bank.”

I remember being flummoxed during my PhD when in my ethics class, the professor admitted to “p-hacking” (checking for statistical significance and then adding more subjects if it wasn’t reached yet), with the defense that “Everyone does it”.

As bad as all this is, the great thing about science is that it can self-correct. By testing past results, we know better now which ones hold up. And there are measures that would make this kind of thing less likely in the future, like researchers being obligated to share the data for the study, and studies being pre-registered with the specific statistical tests they plan to run.

Subscribe now

The Hidden Mind

Read more

Like a good little skeptic, I knew the whole “left brain vs right brain” talk in popular culture was bunk. The idea that some people are “left-brain dominant” and therefore analytical, while others are “right-brain dominant” and therefore creative, has long been called out by scientists as a myth. But here was a neuroscience professor telling me that actually there is a difference between the right and left hemispheres of the brain.

The annoying thing wasn’t that this meant it was true that people who use their left hemisphere more are logical and those that use their right hemisphere are more creative—that’s still malarkey. What was annoying was that this meant the story was more nuanced than I had thought.

This is something I’ve learned is often the case—with any seemingly dumb myth, there’s often some sort of kernel of truth that kicked it off. It would be much more convenient if there wasn’t some grain of truth buried in the mountain of bullshit, but that’s rarely the case with these kinds of persistent neuromyths.

In retrospect, throwing out all claims of lateralization (the technical term for there being a difference between the two sides) was pretty naive. There is an incredibly conspicuous sign of a difference between our brain hemispheres: most humans (around 90%) are right-handed. The right side of the body is controlled by the left hemisphere. Hence, the left hemisphere clearly does something different than the right—the motor skills of our dominant hands (for most of us).

This isn’t the only difference in the hemispheres—more on that later—but it is the easiest to identify. “Handedness” is so easy to identify that, based on fossils, paleontologists think Anomalocaris, a predator that existed 500 million years ago, tended to prefer to hold prey with their left facial appendage. We know this from seeing injury patterns in trilobite fossils (being called trilobites is really fitting here, since it makes them sound like a snackfood). It’s thought Anomalocaris would hold their prey with its left appendage and then flex the right to shake them until their shell cracked.

Lateralization likely evolved multiple times. It’s also fairly common—of 119 different species spread across different branches of the tree of life, only 32% showed no evidence of an asymmetry in limb preference. Animals tend to have dominant hands.

This raises an obvious question: Why would evolution build symmetrical bodies but asymmetric brains?

Why asymmetry is useful

If you think about kids learning motor skills, the advantage of handedness seems obvious: it’s hard enough to learn to use scissors well with one hand, so you’re going to be better at cutting paper if you just put all scissor practice on one side.

The motor skills don’t just automatically transfer over to the other side. Motor learning is use-dependent. If you split practice between both hands, you’re splitting the practice between your two hemispheres, so neither gets as much training.

Humans aren’t the only ones that benefit from this specialization. When fishing for termites with a twig, chimps that only used one hand for this task caught more termites than those who did not specialize. Handedness pays off.

But increased motor skill practice isn’t the only benefit of lateralization. It’s also been suggested that splitting functions across the hemispheres leads to a greater ability to perform two complementary tasks at once.

While foraging, animals typically need to perform two tasks: keeping a look-out for food and a look-out for predators. Experiments suggest some animals literally keep one eye out for predators: both chicks and fish that are lateralized perform better at this sort of dual foraging task than nonlateralized ones. The nonlateralized animals use both eyes for foraging and watching for predators, rapidly switching between the two, and this constant switching lowers their performance in both.

Hemispheric specialization improves practice, and division of labor improves multitasking. But this brings about another mystery: why would a whole population be lateralized in the same direction?

Subscribe now

Why are most of us right-handed?

It isn’t just that individual people favor one side. It’s that the human race is overwhelmingly right-handed. Why wouldn’t there be a more even split?

To make the problem clear, consider that having everyone be lateralized in the same direction carries some costs. If every animal in a population favors the same side, this makes their behavior easier to predict by predators. If you know all fish keep their left eye on watch for predators, approach their right side.

This population-level bias isn’t inevitable, either. There are plenty of animals that show lateralization that do not have a population bias. Individual octopuses show a preference for one eye, but at the population level there is a roughly even split—right-eyed octopuses aren’t more common than left-eyed octopuses.

So why are we humans so biased towards being right-handed? More generally, why are many species asymmetrical at the population level?

Researchers have suggested there’s various advantages for this type of coordination in social species. For example, if all fish in a shoal are biased towards the same direction, when a predator attacks, the dilution effect of turning in the same direction can be more important than the predictability. If everyone turns right, you remain part of the large group and therefore less likely to be eaten.

In species with complex social structures (like humans) there may be other situations where using the same side confers an advantage. When most individuals process communication, use tools, or gesture with the same side, it makes imitation, social learning, and coordination easier.

Why then, if there are advantages to population-level asymmetry, aren’t we all right-handed? Researchers have found that if there’s a push and pull between competitive and cooperative interactions, this sort of minority effect can emerge as an evolutionary stable strategy.

It’s been found in certain competitive sports, like boxing or fencing, being left-handed has an advantage—likely because it’s unexpected and your opponents have less experience competing with someone who relies more on their left hand. While not all species fence (though some penis fence, like bonobos and flatworms), it’s common for social species to physically fight for dominance in various ways.

So if most people are right-handed, there might be a greater cooperative advantage to being right-handed, but as the number of left-handers shrinks, there comes a point where the competitive advantage of being left-handed is large enough that it’s evolutionarily worth it for a minority to go against the grain. The stable point in this push and pull might be a small minority of left-handers and a majority of right-handers.

Subscribe now

Lateralization in humans

When we look at humans, there are a few obvious cases of lateralization. Handedness, as I mentioned, is the most conspicuous. But another case is language—while both sides contribute in different ways to language, in most people (88% of right-handed people, 78% of left-handed), the left hemisphere is more involved in language production.

What are we to make of this lateralization? They can matter clinically—for example, if someone has a stroke, damage to the left hemisphere could leave them unable to produce language, and this has obvious major implications for quality of life.

There are various other differences in the hemispheres, but they tend to show up as statistical biases toward processing in one hemisphere rather than the other, not the enormous clear-cut differences of handedness and language production. There are biases toward right-hemisphere processing of emotional stimuli. The left hemisphere seems to focus more on detail while the right hemisphere is more dominant in noticing larger-scale patterns.

These differences are interesting to neuroscientists trying to tease apart how different functions in the brain are performed, and for surgeons trying to determine the impact of surgery. But for healthy individuals, it doesn’t matter exactly where a brain function happens—to steal a phrase from Jerry Fodor (who was talking more generally about the mind): “[I]t happens somewhere north of the neck. What exactly turns on knowing how far north?” We might similarly ask what exactly turns on knowing if it’s to the left or right.

In everyday life, the two sides of the brain work together. If you’re performing any type of complex task, the two hemispheres are in constant communication with each other. The same can be said of any other division in the brain—most tasks involve widespread networks across the brain. Whether these networks are spread between different lobes or different hemispheres doesn’t much matter for the person doing the task. Although we have a dominant hand, we do not have a dominant hemisphere, so which side a function happens on doesn’t tell you much about the function or the individual.

Despite this, there’s a long history of people being fascinated by the left and right hemisphere differences, playing into all sorts of far-reaching theories. The most famous of these is the idea that creativity resides on the right side and rationality resides on the left.

People care about this left-right dichotomy not because it’s left and right versus top and bottom or front and back. It’s because they see it as lending credibility to a simple psychological model that creativity and rationality are opposed, and different people are either creative or analytical.

Unfortunately, this theory is wrong in multiple ways. First, there are problems with the neuroscience: there’s no clean division where the right hemisphere is creative and the left is analytical. And again, people don’t have a dominant hemisphere, so there isn’t some important division that can be made between individuals.

But perhaps more importantly, the psychological model underlying this theory is wrong. People aren’t either analytical or creative. Some measures of analytical thinking, like IQ tests, find positive correlations with creativity—high analytical ability makes one more likely to score high on creativity test scores. Other measures, like classic tests of analytical-rational versus intuitive-experiential thinking styles, find that they are largely independent, not opposite sides of a spectrum. An individual can be anywhere on either dimension regardless of where they are on the other. There’s no binary classification of some people being creative and others analytical.

There are other attempts to make a fuss about the asymmetries in the brain, like Iain McGilchrist’s speculations that the left hemisphere being dominant explains all of Western civilization’s ills. Like the older mythology tying creative versus analytic modes of thinking to specific hemispheres, these claims don’t line up with the evidence, rely on speculative psychological models with little evidence, and “go far beyond the neurological facts”.

Though the brain does separate some functions in one hemisphere or the other, this doesn’t mean there’s some philosophical importance to what goes on in one hemisphere versus the other. The brain is simply trying to be economical, using the brain matter it has. The lateralization we see historically is tied to situations where fast response times or repetitive motor learning are important. In higher order reasoning, there aren’t such clean divisions—both hemispheres are involved and work together, and little hinges on which half does what.

So there’s more nuance to the whole left-brain right-brain thing than I thought when I first got into neuroscience. There are interesting questions about why we see certain lateralizations and not others. These differences are biologically real and scientifically interesting—but they don’t map easily and cleanly onto grand theories of human behavior.

If you enjoyed this, please hit the “Like” ❤️ button, restack, or share this article to help others find it.

If you enjoy Cognitive Wonderland, consider supporting it by becoming a paid subscriber at whatever level feels comfortable for you.

If you’re a Substack writer and have been enjoying Cognitive Wonderland, consider adding it to your recommendations. I really appreciate the support.

Leave a comment

Share

Subscribe now