How to Get Someone to Open Up: What the Research Actually Shows

First, what this page used to say

This article was published under the title nine phrases that quietly lower anyone's defenses, and its subheading claimed those phrases were the ones therapists use without calling attention to it. It was rewritten in September 2026 because both halves of that were wrong, and the change is logged on our corrections page rather than made quietly.

Lowering someone's defences describes an operation performed on a person who has not agreed to it. Whatever the nine phrases were, the frame was the problem: it cast the reader as an operator and the other person as a lock. And the claim about therapists had no source. Clinical work rests on an agreed relationship with consent and limits, which is close to the opposite of a technique used on somebody who does not know it is being used.

That rewrite fixed the framing but left the page thin — about 2,000 words resting on five references, none of which supported a specific sentence in it. This is the second pass, and it is built the way the rest of the site is built: every empirical claim in what follows has a study behind it in the sources list, the numbers are quoted from the papers, and two claims the earlier version made have been removed for lack of evidence. Those are in the last section, so you can see what changed rather than take our word for it.

The honest answer, before the evidence for it

If you searched for how to get someone to open up, you were probably hoping for a question, a phrase, or a sequence. There isn't one, and that is a finding rather than an evasion. Here is what sixty years of research on self-disclosure does support, stated plainly, with the rest of this article being the evidence.

What you say about yourself matters far less than whether it is a feeling. In the diary study that tested this most directly, disclosing facts about yourself predicted nothing at all about how close an interaction felt. Disclosing an emotion predicted a great deal.

Half of the process is not yours. Feeling that the other person understood, accepted and cared about what you said is what converts disclosure into closeness. That means the most useful thing you can do for somebody else's openness is to be the kind of listener whose reaction is safe to receive.

Inviting someone to talk is not a sacrifice you make to get something later. In the meta-analysis of this literature, one of the three reliable effects is that people end up liking the person they disclosed to. Being the listener is not the losing role.

The shape of the exchange changes the outcome with the content held constant. Same amount of disclosure, taken in turns rather than in one long stretch, produces more attraction.

And the thing stopping most people is a prediction error. People systematically expect deeper conversations to be more awkward, and expect strangers to care less about what they say, than either turns out to be.

Nothing on that list is a technique, and none of it works on a person who does not want to do it with you. What it describes is a two-person process you can enter and invite, and the rest of this piece is about how that differs from pressure — a difference this genre is bad at, which is why our version of it is in the editorial policy.

The finding that matters most: it is feelings, not information

Psychologists have been measuring self-disclosure since Sidney Jourard and Paul Lasakow built a questionnaire for it in 1958, and most of what came after asked how much people disclose. The study that answers the practical question best asked a different one, and it is rarely the one quoted.

Jean-Philippe Laurenceau, Lisa Feldman Barrett and Paula Pietromonaco tested Reis and Shaver's model of intimacy the hard way: not by asking people about their relationships in general, but by having them fill in a record immediately after every social interaction lasting ten minutes or more, for a week in Study 1 and two weeks in Study 2. After each one, they rated how much they had disclosed, how much the other person had, how understood and accepted they felt, and how intimate the interaction was. Study 1 produced 1,494 usable dyadic interactions from 69 participants. Study 2 repeated it with broader measures on 85 people.

Two things came out of it.

Self-disclosure and the other person's disclosure both predicted intimacy at the level of the single conversation — not as a trait of relationships, but interaction by interaction. That is worth stating because most advice treats openness as something a relationship has or lacks, when what was measured is something that happens or does not happen in the next twenty minutes.

And in Study 2, they separated disclosure of facts from disclosure of emotion, and the result is the single most useful number on this page. Controlling for the other person's disclosure, disclosing emotion predicted intimacy at β = .37. Disclosing facts predicted it at β = −.01, which is nothing. The difference between them was large and clearly significant, χ²(1, N = 85) = 98.22, p < .001. Factual disclosure was dropped from the remaining analyses because its path was statistically indistinguishable from zero.

So the version of this that people reach for first — ask better questions, get more information, learn more about them — is not the mechanism. You can learn a great deal of true and private information about a person and generate no closeness whatsoever. What generates it is a feeling, stated as a feeling, and received as one. The practical difference is between I've had three serious relationships and the last one ended in a way I still don't fully understand. Both are personal. Only the second one is doing anything.

The honest limits: this is correlational, so reverse causation — feeling close making you disclose more — cannot be ruled out, and the authors say so. The samples were undergraduates, and the emotional and factual ratings were global rather than coded from transcripts. What it establishes is a within-interaction association, which is a smaller claim than a causal one and still the most directly useful thing anyone has measured about this.

The other half of the process is not yours to do

The same study found the mechanism that explains why this so often fails with a person who is technically saying revealing things.

Reis and Shaver's model has two components, not one: somebody discloses, and the other person responds in a way that conveys understanding, acceptance and care. Laurenceau and colleagues measured the second one — perceived partner responsiveness — and found it partially mediated the effect of both self-disclosure and partner disclosure on intimacy. In Study 2, where responsiveness was measured more broadly (feeling understood, accepted and cared for, rather than accepted alone), its role as a mediator got stronger, and the direct effect of partner disclosure on intimacy dropped from .40 to .32 once it was in the model.

Read that in the direction that matters to you. If you want somebody to open up, most of the variable you control is not what you ask. It is what happens in the two seconds after they answer.

There is experimental work on that half too, and it is more specific than active listening advice usually is. Guy Itzchakov, Kenneth DeMarree and Avraham Kluger ran five experiments, one preregistered, manipulating listening quality in several different ways, and found that merely being listened to well — empathic, attentive, non-judgemental — reduced speakers' social anxiety and led them to explore what they actually thought more deeply. The measured consequence was greater attitude clarity: people became surer of what their own position was, without becoming surer that it was correct. A separate set of studies from the same programme found the same manipulation reduced attitude extremity rather than reinforcing it.

That is the closest thing in the literature to a mechanism for the folk observation that some people are easy to talk to. It is not a set of phrases. It is that a certain quality of attention lowers the cost of the next sentence, and people then say the next sentence.

Three things follow, and all three are things not to do rather than things to do. Do not make your reaction the thing they now have to manage — the fastest way to end a disclosure is to be visibly more upset by it than they are. Do not convert it into a problem statement, because problems get answered and answered conversations stop. And do not treat the moment as the test. What establishes that somebody is safe to talk to is usually the fact that they remembered three weeks later and asked. The same asymmetry runs through the behaviours that get called emotional maturity and through the repair research in how to reconnect after a fight.

Being the listener is not the losing role

There is a widespread intuition that the person doing the talking gets the benefit and the person listening pays for it. The measurements say close to the reverse.

Nancy Collins and Lynn Carol Miller's meta-analysis of this literature identified three distinct disclosure–liking effects, all of them statistically significant: people who disclose intimately are liked more than people who disclose less; people disclose more to those they already like; and — the one that gets left out — people come to like others as a result of having disclosed to them. The three are not competing findings. They describe a loop, and the third one is the part that turns an invitation into something the other person gains from rather than gives up.

Susan Sprecher, Stanislav Treger and Joshua Wondra tested the division of labour directly. Pairs of unacquainted students did a structured disclosure task in which one person disclosed while the other only listened, then swapped roles for a second round. After the first round, the listeners reported more liking and more positive impressions than the disclosers did — and listening was associated with perceiving more similarity. After the swap, the difference disappeared. So over a full exchange it evens out; within any one stretch of it, the person listening is not the one losing.

And there is a reason the talking half feels good too, which is worth knowing if you are the one being invited to go first. Diana Tamir and Jason Mitchell ran five studies on the value people place on talking about themselves. Self-disclosure was associated with increased activation in the mesolimbic dopamine system, including the nucleus accumbens and ventral tegmental area — and, in the behavioural half, people gave up money for the opportunity to disclose about themselves rather than answer questions about someone else. Their opening line is the relevant statistic: humans spend 30–40% of speech output telling other people about their own experience.

Put those together and the picture is not two people trading a scarce resource. It is a thing most people want to do, held back by something else. The next two sections are about what that something else is — first the shape of the exchange, then the fear.

Take turns: the shape changes the outcome with the content held constant

This is the cleanest experimental result on this page, because it manipulates form while holding quantity fixed.

Sprecher and Treger put pairs of strangers through two get-acquainted interactions using a live-interaction paradigm. In one condition the pairs took turns, alternating disclosures back and forth. In the other they went sequentially: one person disclosed at length while the other listened, and they switched roles in the second interaction. Same task, same total amount of disclosure, different shape.

The turn-taking pairs reported greater attraction, and the gap was largest after the first interaction. Three process variables — responsiveness, enjoyment, and the perception of being liked — were also higher in the turn-taking condition, and those three mediated the link between reciprocity and attraction. Which is to say the reason taking turns worked was not the taking of turns. It was that turn-taking produced the feeling of being responded to, and that produced the attraction.

The practical form of this is smaller than it sounds and easy to get wrong in the opposite direction from the one people worry about. The common failure is not talking too much. It is asking a question, receiving an answer, and then asking another question — which feels attentive and reads as an interview, because nothing comes back. Answer your own question, at roughly the depth you are hoping to receive, and do it early rather than after they have gone first.

The version of this for a relationship that already exists is in which questions actually reveal compatibility, where the same turn-taking finding does most of the work. And the version for a first meeting is in first date mistakes: the failure mode there is a person who talks for seventy per cent of the evening and leaves convinced it went well.

What the 36 questions actually found, including the part that got lost

Everything above sits downstream of one 1997 paper, and it is worth going through properly, because the popular version of it has almost nothing to do with what was measured.

Arthur Aron, Edward Melinat, Elaine Aron, Robert Vallone and Renee Bator published a procedure for creating closeness in an experimental context: pairs work through three sets of slips over 45 minutes, fifteen minutes per set, with the questions escalating in intensity within and across sets. That is the list that circulates as the 36 questions that make you fall in love.

Study 1 compared it against small talk, with pairs randomly assigned. On the combined closeness measure — the Inclusion of Other in the Self scale plus the Subjective Closeness Index — the closeness condition averaged 4.06 against 3.25 for small talk, an effect size of 0.88 standard deviations, F(1, 37) = 5.68, p < .05. That is a large effect for forty-five minutes with a stranger, and it is the finding that deserves its reputation.

Now the three things that get dropped.

The authors state that the goal was a temporary feeling of closeness, not a relationship. In their own words the procedure is like a mood induction: useful for creating a state under controlled conditions. They add that it seems unlikely to produce loyalty, dependence or commitment, and that there is minimal shared history in it.

Studies 2 and 3 are a run of null results, with adequate power. Whether the pairs had been matched to avoid disagreeing on important attitudes made no difference. Whether they had been led to expect mutual liking made no difference — in both cases the F ratios were below 1. And in Study 3, whether getting close was set as an explicit goal made no difference either. Wanting it, expecting it, and agreeing with each other were all tested, and none of them moved the outcome. The tasks did.

And the love claim comes from a parenthesis about a different study. Describing their earlier 1991 pilot, in which cross-sex stranger pairs did the tasks for an hour and a half, the authors mention encouraging anecdotal reports over the following months — "including one pair who married!". That exclamation mark is the origin of the entire genre. One couple, in a pilot, reported informally. The word love does not appear as a finding anywhere in the paper.

This is the same pattern this site keeps running into: a real result, correctly reported by its authors, turned into a promise by everyone downstream. The three-day rule and the 3-6-9 rule are the same shape, and so is almost everything said about why smart people repeat their choices.

What survives, and it is not nothing: escalating reciprocal disclosure beat small talk by a large margin, and it did so regardless of what either person wanted from it. The structure did the work.

The reason people don't go first has been measured, and it is wrong

Almost everybody reading this already knows that going first is the move. The barrier is not information. It is a prediction about how it will land, and that prediction has now been tested repeatedly and found to be miscalibrated in a consistent direction.

Michael Kardas, Amit Kumar and Nicholas Epley ran a series of experiments comparing what people expected from deep conversation with what they got. Participants systematically overestimated how awkward deep conversations would be, and the overestimate was larger for deep conversations than for shallow ones. They also underestimated how connected and how happy the conversation would leave them. The reason, isolated across several of the experiments, is specific: people expect others to care less about their disclosures than others actually do. And when that expectation was corrected, participants chose deeper topics.

Two related findings say the same thing from other angles.

The liking gap. Erica Boothby, Gus Cooney, Gillian Sandstrom and Margaret Clark found that after a conversation, people systematically underestimate how much their partner liked them and enjoyed their company. It showed up with strangers in the lab, with first-year students getting to know their dorm mates, and among members of the public at a workshop — and among the dorm mates it persisted for months.

The beautiful mess effect. Anna Bruk, Sabine Scholl and Herbert Kappes ran seven studies on how people evaluate showing vulnerability — confessing romantic feelings, admitting a mistake, asking for help. People rate the same act more positively when somebody else does it than when they imagine doing it themselves. Their explanation is construal: your own vulnerable moment is concrete and full of its specific risks, while somebody else's is abstract and reads as courage.

And the base case: Epley and Schroeder instructed commuters to talk to a stranger. The people who did reported a more positive experience than those who stayed silent — and a separate group of commuters predicted the opposite, partly because they underestimated how interested the stranger would be.

The practical upshot is not be brave. It is narrower and more useful than that: the specific thing you are wrong about is how much the other person wants to hear it. That is a correctable error, and the studies show that correcting it changes what people choose to talk about.

One limit worth stating. These are conversations with strangers and acquaintances, in laboratories, on trains, and in workshops. Nobody has shown the same miscalibration governs the harder case — the partner who has gone quiet over years, where there is history on both sides. It probably still applies in part. It has not been tested there.

Go gradually, but know how thin the evidence for that is

Everybody who writes about this says the same thing: escalate slowly, don't lead with the heaviest item. We think it is right. It is worth being precise about what supports it, because the support is older and thinner than the confident tone of most advice suggests.

The usual citation is Paul Cozby's 1972 experiment, which found a curvilinear relationship between disclosure and liking: a moderate discloser was liked more than either a low one or a high one, and the high discloser was rated as significantly less well-adjusted. Participants reciprocated the intimacy level of the low and medium disclosers, and reciprocity broke down at the high end. That is the empirical root of don't overshare, and here is what it actually was: 30 female participants, in a role-playing experiment, responding to a described other rather than a live person. Collins and Miller's later meta-analysis found the disclosure–liking relationship moderated by study paradigm, disclosure type and the discloser's gender — which is a way of saying it depends, and that the conditions have not been mapped.

So the working rule stands, but state it at its real size. Going further than the current level and seeing what comes back is a reasonable way to run an unmapped gradient. Opening with the heaviest thing you have is not wrong because a study says it fails; it is wrong because it hands the other person something they did not agree to receive and now have to respond to. The feeling that produces is obligation, and obligation is not closeness — it is nearer to what happens when somebody starts assigning stages to a relationship by number.

A note on the word oversharing, since the previous version of this page used it. There is no study showing that people who disclose too much too early are avoiding intimacy rather than seeking it. That claim is in wide circulation, it is psychologically tidy, and it does not appear to have been tested. What Cozby measured was the listener's reaction, not the discloser's motive. We have removed the claim from this article; it is in the last section with the others.

If you press harder, here is what you are actually doing

The most common version of this problem is not a stranger. It is a partner, a friend or a family member who does not say much, and a person who has tried asking, tried asking differently, and is now asking more often.

There is a name for that pattern and a large evidence base on it. Demand/withdraw is the sequence where one person pursues discussion and the other pulls away, which pushes the first to pursue harder. Paul Schrodt, Paul Witt and Jenna Shimkowski pooled 74 studies with 14,255 participants and found a moderate, consistent association between the pattern and worse outcomes overall, r = .360 — higher for relational outcomes (r = .423) and communicative ones (r = .418) than for well-being. The effect was larger in distressed and clinical samples (r = .413) than in non-distressed ones (r = .345), and the two directions were symmetrical: wife-demand/husband-withdraw (r = .380) and husband-demand/wife-withdraw (r = .392) were effectively the same size. It is not a gendered problem, and it is not a communication-skills problem. It is a loop in which each person's move makes the other's more likely.

That symmetry is worth holding on to, because the folk version of this problem is gendered: men do not talk, women want them to. Kathryn Dindia and Mike Allen pooled 205 studies covering 23,702 people and found women disclosed more than men by d = .18 — a small difference, with moderators that matter more than the headline. The gap was larger when the person being told was female, and it was larger to friends, parents and spouses than to strangers. With a stranger, men reported disclosing about as much as women did; only observational measures found a difference. A small average difference is not the explanation for a specific silent person in front of you, and treating it as one is how a pattern that predates the relationship gets read as a fact about men.

It is correlational, so which end starts it is not settled by these data. What the numbers do establish is that pursuing harder is not a neutral thing to try.

Then there is the part that predates you entirely. Mario Mikulincer and Orna Nachshon ran three studies on 352 undergraduates classified by attachment style. Secure and anxious-ambivalent participants disclosed more than avoidant ones. More telling: secure and ambivalent participants disclosed more to, felt better with, and were more attracted to a partner who disclosed a lot — and avoidant participants' disclosure and liking were not affected by the partner's disclosure at all. Going first did not move them. Secure participants also showed more disclosure flexibility and more topical reciprocity, which is the measured version of matching the level in front of you.

Sara Bradford, Judith Feeney and Lydia Campbell found the same thing outside the lab, in 113 dating couples who kept structured diaries of their everyday interactions. Avoidance strongly predicted how much people disclosed. And a second effect, which is the one to sit with if you are the person doing the asking: a partner's anxiety predicted negative evaluations of ordinary interactions — their own and the other person's.

Aron's closeness studies contain the same result by accident. Across 97 pairs, avoidant/dismissive pairs reported less closeness (M = 3.59) than secure pairs (4.10), fearful pairs (4.09) or pairs with a preoccupied partner (4.07) — after running the identical procedure that produced a large effect overall.

So: the least effective available response to somebody who does not disclose is to increase the demand, because for the specific pattern most likely to be producing it, demand is the thing the pattern was built to handle. What is left is unglamorous — consistency without demand, over a longer period than feels reasonable — and that is genuinely hard if your own pattern runs anxious, because silence reads as danger and the impulse is to close the gap immediately. Both sides of that loop are worked through in love or anxious attachment, and the attachment style test is a more useful next step than another script. If the sticking point is what happens when the conversation does start, the conflict style test measures the other half.

And one possibility that this genre almost never offers: some people are private and are going to stay private. That is a compatibility question rather than a problem with a solution, and it belongs with the other honest markers in what a new relationship with real potential looks like.

Where an invitation becomes pressure, and where it stops being either

If you are reading this because somebody close to you has gone quiet, it is worth saying before anything else: a conversation with a therapist or counsellor is a reasonable response to that, both for them and for you, and it is not an escalation.

The boundary here is real and easy to cross with entirely good intentions, so the test has to be behavioural rather than about how you feel while doing it.

An invitation can be declined at no cost. If the other person changes the subject and nothing happens — no visible disappointment, no returning to it later that evening, no cooling in the room — it was an invitation.

Pressure is what remains when declining has a cost. Repeating the question. Explaining why they ought to trust you by now. Naming their reticence as the problem with the relationship. Going quiet until they produce something. All of those can be done kindly, and all of them mean the same thing on the receiving end: there is a price for saying no.

There is also a version of this that is not clumsiness and should be named as what it is. Demanding access to somebody's messages, requiring an account of where they have been, treating privacy as evidence of disloyalty, monitoring who they talk to, or making warmth conditional on disclosure are controlling behaviours. In England and Wales, a pattern of that kind is covered by the offence of controlling or coercive behaviour in an intimate or family relationship, and the Home Office statutory guidance lists monitoring and isolation among its examples. It does not stop being that because the subject is emotional intimacy and the stated motive is closeness. If you recognise it being done to you, the relevant reading is not a communication guide — it is a domestic abuse helpline. If you recognise yourself in it, that is worth taking to a professional rather than to them.

The milder and far more common relative of this is the jealousy-monitoring loop, which runs on the same logic — more information will settle the feeling — and does not work for the same reason. The evidence on that is in the psychology of jealousy.

Not everything has to be said, and the research does not back total transparency

The genre this article sits in tends to treat disclosure as a good with no ceiling: more is better, and anything withheld is a deficit. The measurements are more interesting than that.

Catrin Finkenauer and Harry Hazam asked whether disclosure and secrecy in marriage both contribute to satisfaction, and found that they did — independently. What predicted marital satisfaction was not a person's general disposition to disclose or conceal, which contributed only marginally. It was the contextual measures: what actually happened between these two people. Both disclosure and secrecy, measured that way, contributed.

John Caughlin examined why topic avoidance is so reliably associated with dissatisfaction despite theory saying avoidance can be functional. Across 114 parent–child dyads and 100 dating couples, the association was moderated by people's motivations for avoiding and by characteristics linked to those motivations. Avoiding a topic to protect the relationship is not the same act as avoiding it because the relationship cannot hold it, and the data distinguish them.

Set against that, the cost of the other kind of withholding is real and has been measured directly. Ahmet Uysal and colleagues studied self-concealment from a partner — actively keeping distressing personal information from them — in a survey of 165 people and then in 50 couples keeping daily records for 14 days. Daily self-concealment predicted lower relationship satisfaction and commitment and more conflict the same day, and in lagged analyses, lower relationship well-being the following day. The mediator was thwarted needs for autonomy and relatedness. And a partner's concealment predicted the other person's well-being, not just their own.

So the line is not between sharing and not sharing. It is between choosing not to discuss something and actively concealing something that is weighing on you. The first appears to be compatible with a good relationship and sometimes constitutive of one. The second has a daily, measurable cost that shows up the next day. Which is also why they never talk about anything real and they have not told me one specific thing are different complaints with different answers, and why the honest markers of a relationship going well are behavioural rather than confessional — see what the long follow-up studies measured in couples who lasted.

Doing this by message

This article sits in Texting, so the mechanics deserve a section — with a clear line between what has been measured and what is reasoning.

What has been measured. Richard Slatcher and James Pennebaker took 86 dating couples and had one member of each either write about their deepest thoughts and feelings about the relationship for three days, or write about their daily activities. They then collected the couples' instant messages before and after. Two results: participants who wrote about the relationship were significantly more likely to still be together three months later, and in the days after writing, both they and their partners used more emotion words in their messages — with the increase in positive emotion words partially mediating the effect on stability. Nobody was instructed to change how they texted. The expressive writing moved the register of the messages, and the messages tracked the outcome.

That is a small, specific and quite striking result, and it is consistent with everything above: the thing that moved was emotional language, not quantity of contact.

What is reasoning, not evidence. The standard advice — escalate more slowly by message, and do not raise by text something that needs a response you cannot see — follows from the fact that text strips tone and removes the feedback that tells you whether to go further. We think it is sound. We have not found a study testing it, and we are not going to present it as one.

And what is measurably false. There is no evidence that withholding a reply buys anything. The only experiment on post-date timing found a short wait beat no wait and a long one lost on every measure, which is the subject of the 72-hour rule. The habits that actually damage this are in texting habits that push people away, and the popular claims about what phone behaviour signals turn out to be much weaker than advertised in what phone behaviour on a date actually shows.

What this page used to claim and no longer does

Two claims from the previous version have been removed rather than softened, and both are now on the corrections page.

That removing eye contact makes people open up. The old text said that a lot of difficult conversations happen while walking, driving or washing up, and explained it by saying that removing eye contact lowers the intensity of the exchange — "which is why the car is where families historically say things". The observation may well be true. The explanation was presented as established and is not. The nearest relevant research is Argyle and Dean's 1965 equilibrium account of how people trade off proximity and gaze, and a line of work on gaze aversion under cognitive load, most of it in children answering difficult questions. Neither measured whether adults disclose more with the face removed. Nobody, as far as we can find, has run that experiment. What the gaze literature has measured instead — including the finding that the physiological response to being looked at depends on believing the other person can see you, and disappears when they cannot — is in what eye contact actually predicts.

That oversharing is avoidance in different clothes. The old text called it "frequently the same avoidance wearing different clothes: a way of controlling the interaction by determining what it is about". That is a claim about a discloser's motives, and the disclosure literature measures listeners' reactions. Cozby's curvilinear result, covered above, says a high discloser was liked less and rated as less well-adjusted. It says nothing about why the person was disclosing. Removed.

We are stating both of these at length for the same reason the first section of this article exists. A site that says the evidence is thin in this genre has to apply that to its own back catalogue, and the way to make that checkable is to say what we published, what was wrong with it, and what is left. The standing version of that commitment is the editorial policy, and every change of this kind since August is on the corrections page.