Cognitive science

How to Type Someone Else (and Why You'll Be Wrong)

By PsyCove Editorial · 12 min read · Updated 2026-10-27 · Checked against the sources listed below

Give yourself twenty minutes with a stranger and you will come away with a type for them. The research on how well people read each other is specific about what that guess is worth: your accuracy is not spread evenly across the four letters. It depends on which trait you are reading, and on two of them you are barely better than a coin. Here is what you can see, what you cannot, and where the error gets in.

Key takeaways

Accuracy depends on the trait, not on your skill. In a round-robin study, people were the best judges of their own anxiety, friends were the best judges of their intellect, and every perspective — self, friend, stranger — did equally well on extraversion (Vazire, 2010).
Two of the four letters are close to invisible. Sensing versus intuition and thinking versus feeling are claims about how someone processes. Processing leaves no clean trace on the outside.
A small error does not stay small. When the type indicator was rescored for about 12,000 people, the scores formed one center-weighted hump rather than two clusters (Bess & Harvey, 2002). Most people therefore sit near the cut point, where a few points flips a letter.
The ceiling arrives fast. Judgments made from clips under five minutes predicted outcomes about as well as much longer observation did (Ambady & Rosenthal, 1992).

What Can You Actually See When You Watch Someone?

Watching someone tells you far more about some traits than others, and the split has nothing to do with how hard you are looking. In the study that has become the standard reference on the question, 165 people rated themselves and were rated by four friends and up to four strangers. Everyone then took a battery of behavioral tests, which served as the criterion (Vazire, 2010).

This account would be wrong under two conditions: untrained observers doing no better than chance on every trait, and all three perspectives scoring alike. Neither happened. What follows is a description of how well people read traits, not a diagnostic claim about anyone's mental health.

The results sorted by trait. The self was the best judge of anxiety and mood — the neuroticism cluster. Friends beat both the self and strangers on intellect. On extraversion, all three perspectives did equally well, which means a stranger who had watched you for an hour read that trait as accurately as you did.

That last line is the one to carry into any attempt to type someone. Extraversion is the only letter in the code with a large, cheap, hard-to-fake signal on the outside: how much someone talks, how they look after a long day with people, how much room they take in a group.

Which of the Four Letters Can You Actually Read?

Only one of the four letters is mostly observation. The other three are inferences, and they are not equally good inferences.

The four dichotomies seen from the outside

Letter	What you can see	Share of it that is inference
E / I	Talkativeness, energy after social contact, how much space someone takes in a group	Small. The one letter an observer can read with some confidence.
S / N	Nothing stable. People of both preferences talk in abstractions and in specifics, depending on the topic	Almost all of it — and the inference tends to encode your own preference
T / F	Warmth or bluntness of manner, which is style rather than processing	Most of it. Manners, culture and mood all produce the same surface
J / P	Deadlines met, plans made, bags packed — but only where the situation asks for it	About half. It also overlaps with conscientiousness, a different construct

The S/N row deserves the most suspicion. Sensing and intuition are meant to describe what someone attends to: the concrete data in front of them, or the pattern behind it. From outside, both look like a person talking about something they find interesting. When you decide which one you are seeing, a good part of what you are reading is yourself.

Cronbach named the mechanism seventy years ago — assumed similarity, the share of any judgment of another person that is really a judgment of the judge (Cronbach, 1955).

Someone who reels off detail after detail may be a sensing type who has the facts, or an intuitive type who read about them last week.
Someone who speaks only in themes may be intuitive, or tired, or being polite to a stranger at a wedding.
Someone who finishes your sentences is doing what most fluent conversation does, not demonstrating a cognitive function.

Why Does a Small Misreading Give You the Wrong Type?

A type is produced by cutting a continuous score at a midpoint, so a modest error in reading someone does not stay modest.

The two-humped score distributions that were once taken as proof that people come in kinds turn out to be a scoring artefact. Bess and Harvey rescored the type indicator for roughly 12,000 people with a finer estimation method. The humps flattened into one center-weighted hump (Bess & Harvey, 2002). That matters here for a plain reason: if scores cluster in the middle, then most people are near the cut point, and the cut point is exactly where a small error decides the letter.

The site's own guide to the model's limits treats this as the second of its three structural problems, under the heading of information lost at the cut link. It is a property of the scoring, not a flaw in the person being typed.

Two visible letters do not rescue two invisible ones. It is also why our own result page prints a position on each dimension and keeps the type name as a summary of those positions, rather than the other way round. The summary is the piece a third party reads first, and the piece easiest to over-read.

Do Other People Sometimes Read You Better Than You Do?

Sometimes yes, and the conditions are narrow enough to be useful.

Vazire and Mehl tested the ordinary assumption that you are your own best expert. They recorded what people actually did over four days without asking them, then compared self-ratings and ratings by close others against that record. Close others predicted daily behavior as well as the self did, and the two perspectives often predicted different parts of the same behavior (Vazire & Mehl, 2008).

The same paper opens by measuring how strongly people believe the opposite, so the gap between confidence and accuracy is not a small one. Their conclusion is the sentence to keep.

“There is no single perspective from which a person is known best, and both the self and others possess unique insight into how a person typically behaves.” — Vazire & Mehl (2008)

How Good Are First Impressions, Really?

First impressions carry real signal, and the ceiling on that signal is lower than the confidence that arrives with it.

Ambady and Rosenthal pooled 38 results on predictions made from expressive behavior watched for under five minutes. The average effect size was .39. Observations longer than five minutes did not do better, and it made little difference which channel the judge used — face, voice or body (Ambady & Rosenthal, 1992). Thirty-eight results, one ceiling.

Borkenau and colleagues pressed the same question with 600 participants, each recorded in fifteen short episodes. Impressions taken from thin slices of behavior matched what knowledgeable informants said about the same person, and the returns died quickly: past about six episodes, extra footage bought almost nothing (Borkenau et al., 2004). The strongest single cue for intelligence was how a person read short sentences aloud.

Notice what those studies measured, though. Agreement with informants, and prediction of an intelligence test. Neither is the same thing as getting a four-letter code right.

Where Does the Error Actually Come From?

Four things have to go right before a judgment about another person lands, and each of them can fail on its own.

Funder's model of realistic accuracy names them. The behavior has to happen. You have to be placed to see it.

You have to interpret it correctly. And you have to use that interpretation instead of overriding it with what you already believed (Funder, 1995). A person can be honest, visible and consistent and still get typed wrongly, because the fourth step failed quietly.

The four steps, and how each one fails at the type question

The behavior has to happen. Someone who has never had to plan anything gives you no evidence for J or P.
You have to be there for it. You cannot judge decisiveness in a person you have only ever seen at dinner.
You have to read it correctly. The same flat calm can be patience, exhaustion, or a very good mask.
You have to use the reading. This is the step that most often loses: the belief you already hold arrives first and stays.

John and Robins added the detail that matters most for typing. Judges agreed with each other most on extraversion and least on agreeableness. Agreement rose when a trait was easy to watch and low in evaluation. For the evaluative traits — the ones where a high score sounds like a compliment — agreement with the self dropped below agreement between two outside observers (John & Robins, 1993).

Knowing someone helps, slowly. Paulhus and Bruce followed sixteen groups of previously unacquainted people across seven weekly meetings and found impressions growing more valid as the groups got to know each other (Paulhus & Bruce, 1992). Years of acquaintance buy accuracy. They do not turn a guess into a measurement.

Adding information helps, and so does improving its relevance. Letzring and colleagues put unacquainted students into five conditions that varied how much information judges received and how relevant that information was to personality. Both mattered: judges given more, and better-targeted, material came closer to a criterion built from three kinds of informant — the self, real-life acquaintances, and clinician interviewers (Letzring et al., 2006).

What Happens When You Try to Type the Function Stack?

Typing someone by their function order is a harder problem than typing them by letter, because a function is not a behavior.

Nobody exhibits introverted intuition. What you see is a person talking about a pattern, and then you supply the function. The same sentence can come from Ni, from Ne, or from someone who has read one book about either. A function is a proposed explanation for behavior, not a category of behavior, so it cannot be read off the surface.

“I think in images, not words.” Could be Ni. Could be the way anyone describes a visual memory.
“I argue the other side out of habit.” Could be Ti. Could be three years of debating club.
“I pick up whatever the room is feeling.” Could be Fe. Could be vigilance learned in a difficult household.

Our own test stops where that inference would begin. It measures four dimensions and reports a position on each one. It does not measure functions, and no arrangement of four letters reconstructs a stack — the case for keeping the two claims apart is set out in the review of the eight-function model.

This matters for typing other people because the stack is exactly the part observers claim to see, and it is the least visible part of the model. If what you actually want is the measurable version of the same four dimensions, the guide to the five-factor model sets out what those scores are and how they are built link.

Is It Fair to Type Someone Without Telling Them?

Typing another person is an act with a social cost, and the cost lands on them rather than on you. A guess you keep to yourself is a hypothesis. The same guess said aloud in a meeting becomes a label the person now has to argue with.

The workplace is where this matters most, and the law there is more specific than most people expect.

United States. Employers are the regulated party, not colleagues. The ADA restricts what an employer may ask about disability before a job offer, and a test designed to reveal a mental disorder counts as a medical examination (Karraker v. Rent-A-Center, 411 F.3d 831 (7th Cir. 2005)). Your private opinion is not covered. A hiring decision you influenced may be.
United Kingdom. The Equality Act 2010 asks whether a criterion is a proportionate means of achieving a legitimate aim. A type preference applied to a whole team is a criterion, and it needs an aim.
Everywhere else. Consent is the cheap version of the same rule. Ask before you describe someone back to themselves, and notice that the answer is allowed to be no.

The test is whether the person could hear your sentence without being reduced by it. "You are probably an introvert" can survive that test. "This is why you should not run the project" does not.

So How Should You Handle a Guess About Someone?

Treat a guess about someone's type as a hypothesis with an expiry date, not as information about them.

Ask before you decide. One question — “do you know your type?” — beats an hour of observation, because it uses the one channel with privileged access.
Rate the dimension, not the letter. “How much does this person recharge around people?” is answerable from behavior. “Is this person an E?” needs a cut point you cannot see — which is the whole difference between a trait and a type link.
Say your confidence out loud. Recharging style: fairly sure. Decision style: a guess. Keeping the two apart stops the guess from borrowing credibility from the observation.
Check the base rate. Rare types are rare. If your reading lands on one every time, the reading is the thing being measured.
Let the person overrule you. They have access you do not, on exactly the traits where your access is weakest.

A type you assign to someone else is a claim about your reading, not a fact about them. The four-letter code feels like a description because it is short, and short descriptions are the easiest kind to mistake for complete ones.

Frequently asked questions

Can you type someone else accurately?

Partly, and mostly on one dimension. Observers agree with each other, and with the person, reasonably well on extraversion, and markedly less well on the other three. The published accuracy figures for judgments of others sit around a moderate correlation rather than a match (Ambady & Rosenthal, 1992).

Why do people get their friends' types wrong?

Because two of the four letters are inferences about processing rather than observations of behavior, and because evaluative traits distort the judgment. For traits where a high score sounds like a compliment, agreement between outside observers and the person drops (John & Robins, 1993).

Is it easier to type someone you know well?

Yes, and by less than you would expect. Impressions became more valid as unacquainted groups met weekly for seven weeks, but the improvement was gradual (Paulhus & Bruce, 1992). Familiarity narrows the error. It does not remove it.

Can I type someone from their texts or social media posts?

You can form an impression, and impressions from thin slices are not worthless. But the studies that find signal in small samples are predicting broad outcomes — how talkative someone is, how well they did on a test — rather than sorting a person into a four-letter code.

What is the most common mistake in typing other people?

Letting the letters you can see lend authority to the letters you cannot. A confident read on E/I does not transfer to S/N or T/F. Those two need a question, not another hour of watching.

How to Type Someone Else (and Why You'll Be Wrong)

By PsyCove Editorial · 12 min read · Updated 2026-10-27 · Checked against the sources listed below

Key takeaways

Accuracy depends on the trait, not on your skill. In a round-robin study, people were the best judges of their own anxiety, friends were the best judges of their intellect, and every perspective — self, friend, stranger — did equally well on extraversion (Vazire, 2010).
Two of the four letters are close to invisible. Sensing versus intuition and thinking versus feeling are claims about how someone processes. Processing leaves no clean trace on the outside.
A small error does not stay small. When the type indicator was rescored for about 12,000 people, the scores formed one center-weighted hump rather than two clusters (Bess & Harvey, 2002). Most people therefore sit near the cut point, where a few points flips a letter.
The ceiling arrives fast. Judgments made from clips under five minutes predicted outcomes about as well as much longer observation did (Ambady & Rosenthal, 1992).

What Can You Actually See When You Watch Someone?

Which of the Four Letters Can You Actually Read?

Only one of the four letters is mostly observation. The other three are inferences, and they are not equally good inferences.

The four dichotomies seen from the outside

Letter	What you can see	Share of it that is inference
E / I	Talkativeness, energy after social contact, how much space someone takes in a group	Small. The one letter an observer can read with some confidence.
S / N	Nothing stable. People of both preferences talk in abstractions and in specifics, depending on the topic	Almost all of it — and the inference tends to encode your own preference
T / F	Warmth or bluntness of manner, which is style rather than processing	Most of it. Manners, culture and mood all produce the same surface
J / P	Deadlines met, plans made, bags packed — but only where the situation asks for it	About half. It also overlaps with conscientiousness, a different construct

Cronbach named the mechanism seventy years ago — assumed similarity, the share of any judgment of another person that is really a judgment of the judge (Cronbach, 1955).

Someone who reels off detail after detail may be a sensing type who has the facts, or an intuitive type who read about them last week.
Someone who speaks only in themes may be intuitive, or tired, or being polite to a stranger at a wedding.
Someone who finishes your sentences is doing what most fluent conversation does, not demonstrating a cognitive function.

Why Does a Small Misreading Give You the Wrong Type?

A type is produced by cutting a continuous score at a midpoint, so a modest error in reading someone does not stay modest.

Do Other People Sometimes Read You Better Than You Do?

Sometimes yes, and the conditions are narrow enough to be useful.

The same paper opens by measuring how strongly people believe the opposite, so the gap between confidence and accuracy is not a small one. Their conclusion is the sentence to keep.

“There is no single perspective from which a person is known best, and both the self and others possess unique insight into how a person typically behaves.” — Vazire & Mehl (2008)

How Good Are First Impressions, Really?

First impressions carry real signal, and the ceiling on that signal is lower than the confidence that arrives with it.

Notice what those studies measured, though. Agreement with informants, and prediction of an intelligence test. Neither is the same thing as getting a four-letter code right.

Where Does the Error Actually Come From?

Four things have to go right before a judgment about another person lands, and each of them can fail on its own.

Funder's model of realistic accuracy names them. The behavior has to happen. You have to be placed to see it.

The four steps, and how each one fails at the type question

The behavior has to happen. Someone who has never had to plan anything gives you no evidence for J or P.
You have to be there for it. You cannot judge decisiveness in a person you have only ever seen at dinner.
You have to read it correctly. The same flat calm can be patience, exhaustion, or a very good mask.
You have to use the reading. This is the step that most often loses: the belief you already hold arrives first and stays.

What Happens When You Try to Type the Function Stack?

Typing someone by their function order is a harder problem than typing them by letter, because a function is not a behavior.

“I think in images, not words.” Could be Ni. Could be the way anyone describes a visual memory.
“I argue the other side out of habit.” Could be Ti. Could be three years of debating club.
“I pick up whatever the room is feeling.” Could be Fe. Could be vigilance learned in a difficult household.

Is It Fair to Type Someone Without Telling Them?

The workplace is where this matters most, and the law there is more specific than most people expect.

United States. Employers are the regulated party, not colleagues. The ADA restricts what an employer may ask about disability before a job offer, and a test designed to reveal a mental disorder counts as a medical examination (Karraker v. Rent-A-Center, 411 F.3d 831 (7th Cir. 2005)). Your private opinion is not covered. A hiring decision you influenced may be.
United Kingdom. The Equality Act 2010 asks whether a criterion is a proportionate means of achieving a legitimate aim. A type preference applied to a whole team is a criterion, and it needs an aim.
Everywhere else. Consent is the cheap version of the same rule. Ask before you describe someone back to themselves, and notice that the answer is allowed to be no.

The test is whether the person could hear your sentence without being reduced by it. "You are probably an introvert" can survive that test. "This is why you should not run the project" does not.

So How Should You Handle a Guess About Someone?

Treat a guess about someone's type as a hypothesis with an expiry date, not as information about them.

Ask before you decide. One question — “do you know your type?” — beats an hour of observation, because it uses the one channel with privileged access.
Rate the dimension, not the letter. “How much does this person recharge around people?” is answerable from behavior. “Is this person an E?” needs a cut point you cannot see — which is the whole difference between a trait and a type link.
Say your confidence out loud. Recharging style: fairly sure. Decision style: a guess. Keeping the two apart stops the guess from borrowing credibility from the observation.
Check the base rate. Rare types are rare. If your reading lands on one every time, the reading is the thing being measured.
Let the person overrule you. They have access you do not, on exactly the traits where your access is weakest.

Frequently asked questions

Can you type someone else accurately?

Why do people get their friends' types wrong?

Is it easier to type someone you know well?

Can I type someone from their texts or social media posts?

What is the most common mistake in typing other people?

Letting the letters you can see lend authority to the letters you cannot. A confident read on E/I does not transfer to S/N or T/F. Those two need a question, not another hour of watching.

How to Type Someone Else (and Why You'll Be Wrong)

What Can You Actually See When You Watch Someone?

Which of the Four Letters Can You Actually Read?

Why Does a Small Misreading Give You the Wrong Type?

Do Other People Sometimes Read You Better Than You Do?

How Good Are First Impressions, Really?

Where Does the Error Actually Come From?

What Happens When You Try to Type the Function Stack?

Is It Fair to Type Someone Without Telling Them?

So How Should You Handle a Guess About Someone?

Frequently asked questions

Can you type someone else accurately?

Why do people get their friends' types wrong?

Is it easier to type someone you know well?

Can I type someone from their texts or social media posts?

What is the most common mistake in typing other people?

Further Reading (primary sources & authoritative references)

Related reading

How to Type Someone Else (and Why You'll Be Wrong)

What Can You Actually See When You Watch Someone?

Which of the Four Letters Can You Actually Read?

Why Does a Small Misreading Give You the Wrong Type?

Do Other People Sometimes Read You Better Than You Do?

How Good Are First Impressions, Really?

Where Does the Error Actually Come From?

What Happens When You Try to Type the Function Stack?

Is It Fair to Type Someone Without Telling Them?

So How Should You Handle a Guess About Someone?

Frequently asked questions

Can you type someone else accurately?

Why do people get their friends' types wrong?

Is it easier to type someone you know well?

Can I type someone from their texts or social media posts?

What is the most common mistake in typing other people?

Further Reading (primary sources & authoritative references)

Related reading