The Wall of Other Minds: Perspective-Taking & Perspective-Getting for Understanding Other People & Being Understood

A colleague of mine went quiet in a meeting some years ago, arms crossed, jaw set, and I spent the whole hour reading her. She hated the plan. She thought it was a waste of the budget. She was composing the objection she would deliver the moment the meeting ended, and every gesture I catalogued, the crossed arms, the flat voice when she finally spoke, seemed to confirm it. Afterward I mentioned, carefully, that not everyone was going to be on board, and she laughed. She said the meeting was the best part of her week so far. She had eaten something at lunch that had declared war on her, she had a pain radiating from her jaw to her collarbone, and she had been sitting very still because sitting very still was the only position that helped. I had built a complete inner world for her, furnished it with opinions, given her a plot, and not one piece of it was true. The confidence had felt exactly like knowledge. It was not knowledge. It was fiction, produced by an excellent machine, from almost no evidence.

The last piece in this series ended on an admission: the more vividly you imagine another person's inner world, the more confidently you risk confusing your projection with their reality. Behind the admission stands a fact about the simulator with consequences for every relationship you are in. The simulator is a brilliant generator of hypotheses about other minds and a terrible verifier of them. Imagining another person's interior produces a feeling identical to knowing, and knowing requires something the simulator cannot supply by itself. The most honest use of imagination ends in a question mark.

The wall, and why it is real

Philosophers have a name for the structure of the trouble: the problem of other minds. Every consciousness you will ever meet is experienced from the inside only, and you are inside exactly one. The philosopher Thomas Nagel put the point with a bat. A bat lives its life in sonar, hanging upside down in the dark, finding its dinner in a curtain of reflected sound, and nothing about your ability to imagine hanging upside down in the dark tells you what sonar experience is like from the inside. Nagel's essay, published in 1974 and argued over ever since, was about something bigger than bats. It was about the structure of consciousness itself: there is something it is like to be a bat, something it is like to be your neighbor, something it is like to be the colleague with the crossed arms, and that something is only ever available from one address, the address it lives at.

The wall is real, and most of the time nobody minds, because the wall has doors cut into it by other means. Language is the first: the clumsy, miraculous instrument by which an interior exports itself. Faces are the second, broadcasting constantly whether the owner intends it or not. Time is the third and the largest: years of accumulated knowledge of one particular person, the archive of how she handles bad news, what his silences have meant before, which of her smiles are the real one. With the people closest to you, the doors stand wide, and this is where the simulator acquires both its usefulness and its swagger. A machine that knows your spouse of twenty years reasonably well generalizes that expertise to the stranger in traffic, the new colleague, the quiet teenager, as if fluency in one mind were fluency in minds. Psychologists call the discipline mentalizing, the effort to sense the minds behind the behavior in front of you, a tradition associated with the psychoanalyst Peter Fonagy, and its headline finding can be stated in one line: the sense of understanding another person is a feeling, and feelings can be wrong without lowering their volume.

The overconfidence machine

The machine shows itself in vocabulary before it shows itself anywhere else. Certainty arrives in words it prefers: they always, they clearly, they do not care, delivered on the strength of one observed action. Elizabeth Newton, a doctoral student at Stanford, demonstrated in 1990 how much of this certainty is manufactured. She gave one group of people a list of well-known songs, Happy Birthday, the national anthem, and asked them to tap out the rhythm on a table while a listener tried to name it. The tappers, hearing the full orchestra of the actual song in their heads, predicted that listeners would identify about half of the tunes. The listeners got about one in forty. The gap between the richness carried in one head and the thin signal that reaches another has a name now: the curse of knowledge. Once you can hear the song, it becomes nearly impossible to simulate not hearing it, and you overestimate by an order of magnitude how much of it survives the trip.

The same overconfidence shows up wherever communication is measured. In one well-known study on email, people composed sarcastic or serious messages and predicted how often readers would correctly judge the tone; they expected near-perfect detection and got barely better than a coin flip, while remaining just as confident as before. Nor does deliberating harder fix it. Nicholas Epley and Tal Eyal found that when people try to take another person's perspective, they anchor on their own view and adjust from there, and the adjustment is reliably insufficient: the perspective they end up with is mostly their own, lightly redecorated. The imagine-first strategy does not so much open the wall as paint a door on it.

Perspective-taking can even make things actively worse, in one specific and preventable way. Kenneth Savitsky and his colleagues paired up married couples and had one partner deliver a criticism while the other was coached beforehand to really take the speaker's perspective; some couples also had a spotter, a third party who had listened to the exchange. Perspective-taking raised the receivers' confidence that they had understood the criticism correctly. It did not raise their accuracy, not at all. The spotter, who had actually listened instead of imaginatively emigrating, understood the exchange better than anyone. Imagining another's mind gave people certainty; only attending to what the other person actually said gave them truth.

The machine's confidence has a signature distortion with a name, and it is the most heavily documented bias in social psychology: the fundamental attribution error, the reliable tendency to explain other people's behavior by their character and your own by circumstance. The driver who cuts you off is a reckless fool; you, swerving across two lanes a moment later, had a very good reason. Edward Jones and Victor Harris demonstrated the effect decades ago under conditions designed to remove every excuse: listeners watched a debater argue a position the experimenters had assigned him, knew he had been ordered to argue it, and still inferred his real views from the speech. The simulator skips the circumstance and reaches straight for the soul. It does this because character is a complete story, portable and finished, while circumstance is an unfinished list of things you cannot see. The finished story feels better in the head, and the feeling of completeness is the machine mistaking its own summary for the person.

And the machine has a favorite weather: stress. Under load, attention narrows to the threat, certainty spikes exactly when accuracy collapses, and the practice described here is hardest to reach when it is most needed. Mentalizing is easy when nothing is at stake, which is when it barely matters, and most valuable when your mind is already racing. The countermeasure is preparation rather than willpower: the camera-rivals-feeling-question sequence, rehearsed in calm moments until it is a reflex, survives into the heat far better than any resolution made in the middle of an argument. The practice is not a talent some people have. It is a habit installed in the quiet, spent in the loud.

Where the machine runs hardest

Multiply those experiments by a life and see where the stakes gather, since the overconfidence machine runs hardest exactly where we care most. The marriage argument where each side is certain it knows what the other meant is two simulators locked out of each other's buildings. The parent reading a teenager's silence as defiance has run a story with no evidence, and so has the manager reading a resignation letter as disloyalty, and so, in the other direction, has the adult child reading an aging parent's stubbornness as decline when it may be fear. Beneath every one of these positions sits an interest the certainty never reached: time, money, status, safety, fairness, the need to matter. The teenager's silence protects an interest in being a person apart from the house rules. The parent's stubbornness protects an interest in still counting. In none of these cases does certainty track accuracy, and in every one the cost of being confidently wrong is measured in years. The spotter's lesson scales down to a hallway conversation: in any dispute that matters, prefer the evidence of what was actually said to the testimony of what you are sure they meant. One honest question, asked badly, outperforms a night of expert imagining.

The honest alternative: perspective-getting

There is an alternative, and it has a name in the research: perspective-getting. If imagination is a generator of hypotheses and a poor verifier, the verification has to come from outside, from the other mind itself, the only address where the information lives. Nicholas Epley, who documented the anchoring, is also the one who named the difference: the difference between imagining what your colleague thought of the plan and finding out. Ask. Ask the question, accept the awkwardness, and absorb the answer, which might be stomach trouble and a good meeting. Getting a perspective is slower than taking one, because it requires a conversation, and it requires surrendering a finished story for a live one.

The craft of it is smaller than people fear. The questions that work are open and honest, built to find out rather than to confirm: what did you make of that, what was it like for you, what am I missing? The questions that fail are disguised verdicts, the why would you, how could you family, which carry the accusation inside them and invite defense instead of disclosure. The trap words are small and lethal: right?, obviously, so you're saying, each one closing the door it pretends to open.

There is also a way to hold the hypothesis out loud without starting a fight, and it is worth practicing because the alternative, stating your reading as fact, is how most attempts at honesty go wrong. Label the inference as an inference. I may be reading this wrong, and here is what I think I am seeing, is a sentence that invites correction instead of defense. So is the version that asks to be corrected outright: can you tell me if that is close? Presenting your guess as a guess is the cheapest verification there is, and it carries a side benefit the certainty never offers: people explain themselves more openly to someone whose mind is visibly still open.

Understanding and agreement are different acts, and the mind's habit of fusing them is the reason most people underuse the whole practice. The mind resists taking a rival's perspective on the quiet suspicion that grasping the position means conceding it, and it resists mentalizing people it loves on the opposite suspicion that being understood means being agreed with. Neither suspicion survives contact with the actual skill. You can hold the whole of another person's reasoning, state it better than they stated it, and still decline it, and the declining gets firmer afterward, since you now know what you are declining. The skill is also honest about its own amorality: con artists and interrogators mentalize brilliantly, which is one reason the practice belongs to a person of character. Grasping a mind and endorsing a mind are different acts, and confusing them costs you both truth and company.

One worked example, since the asking sounds easier than it is. A friend was sure his brother was furious with him: one-word texts for two weeks, a canceled dinner, and a simulator that had built the whole case, the grudge, the grievance, the speech being saved for Thanksgiving. He almost wrote the apology preemptively. Instead he asked, plainly, with the question that finds out rather than confirms: you have seemed far away, is everything all right? The answer rearranged everything. The brother was not angry at him; he was ashamed at work, had told no one, and the canceled dinner was hiding a man who could not stand to be seen. The one-word texts were self-protection, aimed at the world. Two simulators, two wrong stories, one question, and the four-state reception of the answer: the flatness in the texts was depletion, the belief underneath was that shame must be soloed, the feeling then was dread and the feeling now is relief, and the practical fact is a workplace failure he has since described out loud. The imagined grudge was vivid. It was also, like most vivid things the simulator produces about other minds, entirely free of charge to make and expensive to keep.

And the receiving end is already practiced: the borrowed-life practice from the last piece is this practice under another name. An answer arrives as an experience, yours now, and the discipline of taking it through all four states, what it cost them in energy, what they believed, what they felt then and what they make of it now, what actually happened, is exactly how an answer gets received as knowledge instead of skimmed as news. The question opens the door, the four-state reception walks through it.

There is a second gift in the asking, less obvious and worth naming. People are not transparent even to themselves, and a well-asked question gives them the occasion to find out what they already half-know. The colleague who went silent knew she felt strange, and only the asking produced the discovery that strange meant lunch. Waiting for people to volunteer their inner worlds rarely works, since most people err in the opposite direction, overestimating how visible their distress is, assuming their feelings are written plainly on them when they are sealed inside. Psychologists call it the illusion of transparency, and it means both sides of the wall are misreading at once: you are inventing their interior, and they are overestimating what you can see of yours. Emily Pronin's research adds a final asymmetry to the traffic: people believe they know others better than others know them, in every direction, everyone grading their own insight more generously than anyone else's. The question is the correction on every end.

If it appeals to you, the whole discipline compresses into a moment small enough to use this week. The next time certainty about another person arrives quickly, and it will, pause long enough to do three things. Write the action the way a camera would have recorded it, no adverbs, no motives, only what a lens could capture: she left the meeting after ten minutes. Then let the simulator do the one thing it is actually qualified to do, and generate not one explanation but three, and make one of them unrelated to you, since the story that centers you is the one your machinery finds most flattering and least reliable. And notice, while you are at it, which feeling has been lobbying for the first story, since fear makes silence feel like rejection and anger makes a mistake look deliberate, and a feeling named stops masquerading as evidence. Then ask the one question that could dislodge your favorite explanation. And when the answer comes, say it back to them and ask what you are still missing, since the loop closes only when the other mind has corrected the story it heard itself entered into. Camera, rivals, feeling, question, check. It fits between two breaths, and it is the difference between responding to a person and responding to your fiction about them.

What asking cannot do

Honesty requires the other half of the ledger, and it fits here naturally. Asking has its own limits, and pretending otherwise would be the same overconfidence in different clothes. Some questions are intrusions, and part of the craft is reading which doors are yours to knock on; people have the right to an unreadable interior, and a friend who demands transparency in the name of honesty has traded one wall for another. Some answers will be wrong, since people narrate themselves in the terms available to them and those terms lag behind what they actually feel. And some of what matters never reaches language at all, which is why knowing a person over years still outranks any single conversation. Asking is not the end of the wall. It is the only door in it, and it only opens partway, and it works so much better than the alternatives that the partialness stops mattering.

There is also a mirror mistake waiting on the other side, and it deserves equal time. The first failure was treating your interpretation as fact. The opposite failure is pretending behavior means nothing, or postponing every decision until the motive is finally known. You will often never know it. People go quiet without reasons they could name, leave situations without explaining, and the absence of certainty is never the absence of a decision: the silence itself is a fact you can act on. Curiosity is not paralysis, and a boundary drawn on what was actually said and done is not a failure to mentalize. It is the discipline done properly: hold the intention as a hypothesis, and act clearly anyway.

The whole method

The final position is double, and each half earns its keep. Imagination keeps its job, the one it does brilliantly: proposing. It generates the hypotheses, runs the scenarios, imagines what the silence might mean, so that the questions are good ones. Where it was never qualified was verification, and now it has a colleague for that. The simulator proposes, the other person disposes, and the sequence of the two is the whole method: imagine first, ask second, receive through all four states, and notice, on every round, that the confidence you feel in your reading of another person is data about you and about nothing else. The colleague with the crossed arms taught me that, and it cost her a lunch I have never let myself forget.

Being understood is only half of this craft, and the other half, making sure your own mind arrives with less loss than the tappers' songs, is a skill with its own research and its own practice. The last piece in this series takes up making things stick, in memory, in habit, in the self you are becoming, and it goes there carrying everything the wall taught: that knowing your own interior is as effortful as knowing anyone else's, and worth the same kind of work. My colleague and I laugh about the lunch now. She has crossed her arms in meetings many times since, and each time I notice the thought arrive, fully furnished, and I let it stay a hypothesis until I have asked. That is the whole practice. It is smaller than the wall, and it goes through it.