TOPIC 21 · PHASE C SELF-MODULATION

Awe & the Overview Effect

In the second that stops you, what changes isn't a thought — it's the shelves

2026-07-28 · BigCat

Awe isn't being moved. It's your model of the world admitting, on the spot, that it isn't big enough.

You've probably had the second: a deep-field photo, or the whole Milky Way overhead on a dark night, or standing under a tree that is unreasonably large. Nothing to say, goosebumps down your back, and for that second you weren't thinking about anything. Most people file it under "nice view" and move on. Emotion science pulled that second out for separate study, because it's the one emotion whose trigger condition is that your current understanding can't hold what's in front of you. And here's the part that matters: while it lasts, the projector that runs your personal story all day goes quiet — the thing last issue took weeks of practice to nudge, a tree hands you for free.

// 01

Two conditions: too vast, and it doesn't fit

In 2003 Keltner and Haidt took a word that had been floating around religion, aesthetics and philosophy and reduced it, for the first time, to something you can operationalize. Every clear case of awe meets two conditions at once:

Perceived vastness — the thing is on a scale outside what you normally deal with. And a need for accommodation — your existing framework can't absorb it, so the framework has to change.

The second condition is the one that gets skipped, and it's the one that does the work. "Accommodation" is borrowed from developmental psychology, where it pairs with "assimilation." Assimilation is when a new thing has a shelf waiting for it (another dog you haven't seen before goes on the "dog" shelf; done). Accommodation is when the shelves themselves won't do, and you have to rebuild them.

ASSIMILATION it fits a shelf is already waiting too big ACCOMMODATION it doesn't the shelves get rebuilt
Awe = vastness + need for accommodation: not shelving a new thing, but rebuilding the shelf

That separates awe cleanly from every other emotion. Fear, anger and joy appraise an event — is this good or bad for me. Awe doesn't appraise the event. It flags your model itself: this worldview of yours can't hold what's in front of you. Back on the thread from Topic 1, awe is a prediction error too large to quietly smooth over — so large the only move left is to change the model.

Worth noting: vastness need not be physical. A proof that dissolves a problem you'd been stuck on for three days, a chorus that flattens you out of nowhere, someone doing the right thing when nothing forced them to — all trigger the same machinery. Keltner and Haidt also listed five appraisals that flavour awe: threat, beauty, exceptional ability, virtue, and the supernatural. The first one gets its own section later.

AI cross-read

This one stings a little. Assimilation is explaining new input with knowledge you already have; accommodation is genuinely revising the internal model. Today's AI models, while talking to you, can only assimilate — never accommodate: the weights are frozen, and there's no way to redraw the shelves mid-conversation. So when something genuinely exceeds a model, it can't register "this doesn't fit here." It jams the thing onto the nearest shelf and confidently invents the details — one real source of hallucination. Humans have a dedicated emotional signal for "this is beyond me." A model has none: it doesn't know what it doesn't know. Giving a model a stable sense of its own ignorance — the field calls it confidence calibration — is, structurally, installing an awe alarm.

// 02

For that second, the projector running "my story" gets turned down

If awe really is doing something to the model, something should be visible in the brain. van Elk (2019) put people in a scanner with three kinds of video — awe-eliciting (aerial sweeps of mountains, cosmic-scale shots), merely pleasant, and neutral — and asked them either to become fully absorbed or to count how many times the camera angle changed.

The result landed on a network we already know well: the default mode network, the system that's busiest when you're daydreaming and whose job is running the movie of your own story (medial prefrontal cortex, posterior cingulate, angular gyrus). Absorption normally drives this network up. During awe videos, that rise was noticeably suppressed. Full immersion, without thinking about yourself — which is not what most absorbed states look like (default mode network).

The behavioural side gives you the same thing from the other end, under the name small self. Piff (2015) tested it across five studies and more than two thousand participants: have someone stand for one minute looking up into a grove of absurdly tall trees, and afterwards they're more likely to help a stranger pick up dropped pens and less likely to act as though they're owed special treatment. The variable that statistically linked the two was self-reported smallness.

The most charming evidence needs almost no equipment. Sturm (2020) had older adults take a 15-minute walk once a week; the only difference between groups was one extra instruction — "on the way, deliberately look for things that amaze you." Each walk ended with a selfie. Over eight weeks, it wasn't only their mood ratings that improved. The photos changed: they got smaller in the frame, the world behind them got bigger, and their smiles got wider.

WEEK 1 8 weeks WEEK 8
Selfies after eight weeks of awe walks: the self shrinks in frame, the world grows, the smile widens

Stack those three findings and they point at one move: awe doesn't make you feel good so much as temporarily turn down the weight on the variable called "me." Last issue's meditation gets that network quieter only through repeated practice; standing under a thousand-year-old tree, it goes quiet on its own. Which is why awe keeps getting discussed alongside ego dissolution (Topic 17).

// 03

The overview effect: the same thing pushed to the human limit

In 1987 a writer named Frank White gave a name to something astronauts kept describing: the overview effect. Seen from orbit, Earth hits people with something they struggle to put into words — it has an edge, that thin smear of atmosphere is the entire life-support system, national borders simply aren't on it, and everyone you have ever known is on that ball right now.

Yaden (2016) worked systematically through astronaut reports scattered across memoirs and interviews, doing something plain but essential: translating a heavily romanticized experience back into psychological constructs that already exist. The translation showed nothing mysterious. It's awe, plus a family of states called self-transcendent experience: decreased salience of the self, increased sense of connection to something larger.

Yaden (2017) then laid that family out along a single continuum. On one axis you get everyday mindfulness and flow, then awe, then the overview effect, then full mystical experience — a difference not of kind but of degree: the further right, the fainter the boundary around "me" and the stronger the connectedness.

mind-wandering mindfulness · flow awe overview effect mystical salience of "me" ↓ connectedness ↑ degrees along one axis — not five separate things SELF-TRANSCENDENT EXPERIENCE · ONE CONTINUUM
From mind-wandering to mystical experience: the degree changes, not the kind

That translation carries a very practical conclusion: what does the work isn't space, it's a stimulus that maxes out both vastness and need for accommodation at once. Astronauts hit the extreme values of those two conditions; they didn't visit a place with magic in it. So domes, VR, even aerial footage do push in the same direction on the ground — van Elk's participants were watching video in a scanner and the default-mode effect showed up — just much weaker.

The other half of the honest account: the astronaut evidence is retrospective self-report — recalled after the fact, a few dozen people, and the most ruthlessly selected humans on the planet. Suggestive, but not to be used as causal evidence.

// 04

What the body does, and where the evidence stops

Awe is one of the few emotions with an unmistakable bodily marker: goosebumps. The move itself is useless in humans (we have no fur left to raise), but the autonomic circuitry that drives it is intact — the system you don't consciously control, the one that sets heart rate, breathing and digestion (the hypothalamus is its main command post).

Gordon (2017) found a crucial fork here: awe has two variants, and their bodily signatures run in opposite directions. The kind you get from a starry sky, a huge tree, a piece of music — positive awe — comes with parasympathetic activation: heart rate slows, the body opens, the posture is approach-and-explore. The kind you get from a tornado, a terrorist attack, a wrathful god — threat-based awe — runs sympathetic: heart rate climbs, real fear is mixed in, and participants report low control over the situation (amygdala). The consequential difference is downstream: positive awe raises momentary well-being, threat-based awe does not.

same trigger: vast + doesn't fit positive awe stars · tall trees · a chorus parasympathetic · heart slows open, approach, explore threat-based awe tornado · disaster · wrathful god sympathetic · heart climbs fear, low control, no well-being lift
One emotion, two variants — and the body responds in opposite directions

So "go find more awe" needs a qualifier: the threatening half doesn't count. It also explains why disaster news — equally vast, equally hard to comprehend — mostly just leaves you depleted.

Evidence that goes deeper into the body should be read carefully. Stellar (2015) asked a couple of hundred young adults which positive emotions they'd felt that day, then drew blood for inflammatory markers: awe was the strongest predictor of lower IL-6 (a pro-inflammatory signalling molecule) of any positive emotion. It gets cited constantly, but it's cross-sectional, self-reported, and not large. A lead worth chasing, not a causal conclusion.

The finding that best captures the state of this field is one that fell over. Rudd (2012) famously reported that awe makes people feel they have more time, and it went everywhere. In 2019 van Elk and Rotteveel ran two studies with stricter time-perception tasks and could not replicate it. The interesting part: that's the same researcher who produced the elegant default-mode result. Getting a positive finding and then personally knocking down another claim in your own field is the attitude most worth remembering here.

Monroy and Keltner (2023) organize the existing results into a model with five pathways (neurophysiology, reduced self-focus, prosociality, social integration, meaning). Worth reading — but as a framework awaiting testing, not a conclusion.

One practical note on dose: you don't need space, and you don't need to wait for a trip. The only handle with randomized support is the smallest one — a 15-minute walk once a week, plus the instruction to deliberately look for what amazes you. In diary studies, most reported awe comes from clouds, tree canopies, music and other people's kindness, not the Grand Canyon.

🌀 CROSSING OVER · interdisciplinary leaps

// DEEPER QUESTIONS

If awe depends on "it doesn't fit," does it wear off with repetition?
It does — and that's exactly where the model earns its keep. The twentieth time you see the same sky, your shelves have already been rebuilt: prediction error goes to zero, and the jolt doesn't come. So staring at the same deep-field wallpaper every day is doomed, while rotating everyday small awe (today's clouds, tomorrow an unfamiliar chorus, the day after an insect's mouthparts) holds up. It also yields a counterintuitive corollary: the more you know, the more awe is available, because you can see more of the edges where your understanding runs out. Ignorance isn't full of wonder; it just never notices that the shelves have an edge.
The small self and depressive self-devaluation both shrink the self — same thing?
No — opposite directions. The small self is the weight on "me" dropping: I matter less, attention frees up for other things, and connectedness and prosocial behaviour rise with it. Depressive self-devaluation is the weight on "me" rising while the valuation goes negative: nothing but yourself in your head, all of it bad, accompanied by withdrawal and isolation. It maps onto the default mode network too: depression often shows that network stuck in rumination, elevated and sticky, whereas awe quiets it. One line: the small self says "I'm not that important," depression says "I'm terrible" — the second is more self, not less.
Can the overview effect be manufactured? Could VR substitute for actually going up?
Partly. Since the active ingredients are vastness and non-fit rather than weightlessness or vacuum, in principle any sufficiently maxed-out stimulus pushes in that direction — scanner video already produces a measurable default-mode effect. But ground versions are clearly weaker, and what's missing probably isn't pixel count. It's three things: your body actually being there (vestibular input, weightlessness, a visible curvature), irrevocability (you can't take the headset off), and cost (a view bought with years of training is not the same object in the brain as one you can replay at will). If VR wants to close the gap, the thing to add is probably "no exit," not resolution.
Awe makes people more helpful — or is that just "being in a good mood"?
This is exactly the right thing to press on. Piff's controls included both neutral states and other positive emotions (plain amusement, for instance), and the mediation analysis pointed at the specific variable of feeling small rather than general positive arousal — that's why it's more convincing than "good mood." But emotion-induction experiments generally suffer from small effects and unstable induction strength, and independent replications aren't large enough yet. The safe statement: the mechanism is coherent, the direction has been repeated, the size is undetermined.
Could an AI have awe — and would it be worth building?
Split it in two. First, detection: you need a signal that says "this input exceeds my model." That's precisely one of the capabilities today's models most lack — they sound equally certain about things they've got and things they haven't. Second, what follows detection: a person in awe lowers the weight on self, becomes more cooperative, and actually revises the model; the machine equivalent is "admit uncertainty → go get evidence → update." So the thing worth building isn't making it "feel stunned" — it's wiring up that alarm line. A system that knows when it isn't good enough is a great deal safer than one that is always confident.

// FURTHER READING