attraction.lab

How Many Dates Before You Know? What the Research Can Actually Time

Nobody has counted. In the one study that plotted whole relationships, the ones that lasted and the ones that did not looked identical through the first kiss.

Research-backed writing on attraction, dating and relationships. Drafted by AI against a named bibliography, published only when every claim links to its source.
23 min read
Editorial illustration for: How Many Dates Before You Know? What the Research Can Actually Time

Before the article. A date count is the shape pressure takes when somebody wants to apply it, and the search results for this question sit next to results about how many dates before sex. No number of dates obliges anyone to feel certain, to become exclusive, or to do anything physical, and pressure that continues after a no is coercion rather than impatience. In the United States the National Sexual Assault Hotline is 1-800-656-4673 and the National Domestic Violence Hotline is 1-800-799-7233, or text START to 88788. Find A Helpline, in the sources below, lists free and confidential services by country.

With that said. The honest answer to this question is that nobody has counted. Two literature searches run for this article returned no study that times the point at which a person comes to know, and the numbers in circulation — three to five, two or three, ten — trace back to a dating app's in-house expert, a lifestyle column and an AI summary citing a dating company's advice pages.

What does exist is better. One study cut five-minute conversations into slices and measured how much exposure each kind of judgment needs, and found that the trait people most want to read in a partner is the one read least accurately. Another plotted whole relationships event by event, from the first meeting to the end, and found that through the coffee date, the dinner date, the declaration and the first kiss, the relationships that lasted and the relationships that did not were indistinguishable in how much romantic interest people reported feeling. A third counted how relationships actually start and found that two thirds of them did not start with dates at all, but with a friendship that lasted a median of a year.

Twelve sections: the pressure problem, where the number comes from, the literature counted, what one meeting can be read for, the four-minute result, the study that plotted whole relationships, the table where the two look identical, where they finally separate, why "short-term" is mostly a label applied in hindsight, the two thirds who never had a date count, what early dates are genuinely good for, and the limits. Our editorial policy explains why every empirical claim here carries a linked source.

#1

Before the rest of this page: when a number becomes a schedule

Most people asking this question are asking it in good faith, and usually about themselves: am I taking too long to decide, is it too early to feel sure, is three dates enough to stop replying. That is an ordinary question and the rest of this page is about what the evidence can and cannot say about it.

It needs one thing said first, because of the company the question keeps. In the search results measured for this page, how many dates before you know sits a few positions away from how many dates before sex and how many dates before going to his place, and a date count is the exact form that pressure takes when somebody wants to apply it. A number turns into a ledger very easily, and the sentence it produces — we have been on four dates — is the sentence of someone who believes a decision has been accrued.

Nothing has been accrued. There is no quantity of dinners that obliges anybody to feel certain, to become exclusive, to be kissed or to have sex, and a partner who treats a count as a debt is not negotiating a timeline. Pressure that continues after a no is coercion, not impatience, and it is not a disagreement about pace.

In the United States the National Sexual Assault Hotline is 1-800-656-4673, and the National Domestic Violence Hotline is 1-800-799-7233, or text START to 88788. Find A Helpline, in the sources at the foot of this page, lists free and confidential services by country.

With that said: the honest answer to the question is that nobody has counted dates, and that the research which has timed something is more interesting than the number would have been.

#2

What number is actually in circulation, and who is behind it

The first page of results for this question was pulled on 3 November 2026, from Google with `hl=en&gl=us` and from DuckDuckGo, and every page that answers with a figure was opened rather than judged by its title.

What ranks: a Psychology Today blog post from 21 October 2025, YourTango, Empathi, wikiHow, Live Bold and Bloom, marriage.com, Regain, an Instagram reel, and threads from Reddit, Quora and Facebook. The numbers they give are three to five, two or three, ten, and — in two cases — no number at all.

The Psychology Today piece is the strongest page on the list and is worth describing precisely, because it is the one a careful reader would trust. It answers three to five dates. It cites no study, names no researcher, and reports no sample size anywhere in the article. The figure is attributed to a quote from a person described as the resident dating expert at a dating app. The reasoning offered around it is a set of sayings: once is an accident, twice is a coincidence, three times is a pattern.

Above all of them, Google's own AI summary assembles a staged timeline — one to two dates as a screen, three as a checkpoint, five or six for the connection phase — and attributes those stages to a dating company's advice pages and a YouTube channel.

This is a pattern the site has now documented several times. The three-day rule came out of a 1996 film in which the characters invent the number at the table and cannot agree on it. The 3-6-9 rule and its numbered cousins have no origin at all. The difference here is smaller and worth being fair about: three-to-five is not a fabricated rule so much as a reasonable person's hunch, repeated until it acquired the texture of a finding. Nobody measured it.

#3

The literature, counted on 3 November 2026

Here is the base rate for the question itself, with the queries written out so the count can be re-run rather than taken on trust.

Searching Europe PMC for `"how many dates" AND (romantic OR dating)` returns 8 records. None of them is about this question: they are about relationship functioning in first-episode psychosis, video quality of experience, young women's talk about heterosexual relationships, and the chronology of North American copper. Broadening to `"number of dates" AND romantic` returns 24 records, and again none is about when a person comes to know. The closest are a daily diary study of dating in social anxiety disorder and a blind-date study of age preferences, where the number of dates is a covariate rather than the subject.

So there is no study that counts dates to the point of knowing, and this page is not going to invent one. What there is instead is a small set of studies that timed the thing the question is pointing at — how much exposure buys how much accuracy, and what the early weeks of a relationship look like when you plot them against the weeks that follow. Those have answers, and the answers are better than a number would have been, for a reason that only becomes clear at the end.

One lesson from running this count is worth passing on, because it applies to anyone checking a claim like this: a single database is not a literature search. Europe PMC indexes more widely than PubMed and produced records PubMed does not, and the pattern has caught this site out before.

#4

What one meeting tells you depends entirely on what you are trying to read

The best-designed study of how much exposure a first impression needs is not about dating at all, which is precisely why it is useful here: nobody was trying to make it produce advice.

Dana Carney, C. Randall Colvin and Judith Hall videotaped 30 people in five-minute conversations with a partner and collected criterion scores for each of them — self and peer reports, plus a measured intelligence score. Then they cut the tapes into slices and showed them to 334 judges in a design crossing four slice lengths (5, 20, 45 and 60 seconds) with three locations in the conversation (the first, third and fifth minute), plus a condition that saw the whole five minutes. Accuracy was defined plainly: the correlation between what a judge said about a target and that target's criterion score on the same thing.

The headline is that accuracy improved with exposure, that later segments of a conversation were read more accurately than earlier ones, and that 60 seconds gave the best ratio of accuracy to length. But the useful result is the breakdown by what was being judged. Averaged across the study, judges read extraversion at r = .42, negative affect at .32, conscientiousness at .28, intelligence at .22, neuroticism at .21, positive affect at .20, openness at .17 — and agreeableness at r = .11.

Now read that list as a person on a first date. The thing that comes through fastest and most accurately is how outgoing somebody is. The thing that barely comes through at all is agreeableness: whether a person is warm, considerate, cooperative, easy to be around when something goes wrong. Negative affect, extraversion, conscientiousness and intelligence were judged moderately well from five seconds. Positive affect, neuroticism, openness and agreeableness needed more time to reach the same level.

So the honest version of you can tell a lot from one meeting is that you can tell a lot about the traits that make somebody good at a first meeting. Two limits worth keeping: these are judges watching strangers on video, not people on dates, and the full five-minute condition rested on only 24 of the 334 judges.

#5

Four minutes, and the thing it still could not reach

There is one well-known result that looks like it should settle this, and it is worth saying exactly how far it goes before moving on.

Samantha Joel, Paul Eastwick and Eli Finkel ran more than a hundred advance measures on 350 speed daters and then let machine learning try to predict who wanted whom after four-minute dates. Fed everything collected beforehand, the models predicted essentially none of the variance in who clicked with whom specifically. Fed the measures collected after the four minutes, they predicted 16% to 29% of it.

That is sometimes quoted as evidence that four minutes is enough. It is not what it shows. It shows that four minutes of contact generates information that a questionnaire cannot, which is a statement about the gap between a profile and a person — and it is the reason stated preferences stop working the moment you meet. The full version of that argument, including the 11,196-couple study that puts a ceiling on what any of this predicts, is in what questions can and cannot reveal about compatibility, and the matching-algorithm version of the same problem is in whether dating apps actually work. This page is after something narrower: not what can be predicted, but when.

#6

The study that plotted entire relationships from the first meeting

To answer a question about timing you need data with time in it, and there is one body of work that went and got it.

Paul Eastwick, Elizabeth Keneski, Taylor Morgan, Meagan McDonald and Sabrina Huang adapted a relationship-reconstruction technique to compare people's real short-term and long-term relationships across their whole span. Participants were shown a list of 48 common relationship events — first met, first spent time one-on-one, first short date, first long date, first flirted, first kiss, first told friends, first met the parents, and so on — marked which ones had happened and approximately when, then rated, on a 0 to 100 scale, how much romantic interest they felt at each one. They saw their own trajectory plotted on a graph and could go back and correct it.

Five studies. Study 1 was 86 undergraduates, 70 of them women, each reporting on both a short-term and a long-term relationship. Studies 2a and 2b replicated it with older crowdsourced adults — 143 and 187 people, with 2b preregistered. Aggregated across the set, the analyses rest on 257 ended long-term relationships and 291 ended short-term ones.

Two design decisions deserve credit, because they are the ones a weaker study would have skipped. Each participant reported on both kinds of relationship, so the comparison is within the same person rather than between different people. And the authors analysed only ended relationships, having noticed that short-term relationships are nearly always over by the time anyone reports on them while long-term ones are often ongoing — a difference that would otherwise have manufactured most of the result on its own.

Further illustration for: How Many Dates Before You Know? What the Research Can Actually Time
#7

Through the coffee, the dinner, the declaration and the first kiss, the two look the same

Here is the table that answers the question, with the long-term figure first and the short-term figure second, both on the 0 to 100 romantic-interest scale.

First met the person: 34.2 against 41.0. First spent time together one-on-one: 61.7 against 63.6. First went out together in a group: 57.1 against 52.6. First flirted: 53.4 against 59.8. First went on a short date, the coffee or drinks: 69.8 against 66.4. First went on a long date, the dinner or the film: 77.9 against 74.7. First told the person you were romantically interested: 76.9 against 76.6. First held hands: 73.7 against 70.6. First kiss: 78.2 against 75.0.

None of those differences was statistically significant. Read the list again and notice what it covers: the whole of the period the original question is about. The coffee date, the dinner date, saying it out loud, the first kiss. At every one of those points, the relationships that turned out to be long-term and the relationships that turned out to be short-term were, in how much romantic interest the person reported feeling, indistinguishable.

If anything, the short-term ones started marginally higher — 41.0 against 34.2 at the first meeting, and 59.8 against 53.4 at the first flirtation. The difference is not significant and should not be made into a moral, but it is the opposite of the folk theory, which holds that the ones that last are the ones that felt like something from the start.

The divergence, when it comes, arrives at the sexual events: 92.0 against 73.5 at first intercourse, 82.6 against 69.8 at first oral sex. By then, in the order participants reported, roughly twenty events have already happened.

#8

Where they finally separate, and it is months rather than dates

The authors put the same pattern in effect sizes, and this is the part that converts the finding into an answer about time.

Across the first ten events of a relationship, the differences between short-term and long-term relationships on the behaviours that matter — desire to self-disclose, desire to receive self-disclosure and two related measures — were d = .18, .19, .06 and .06. For events eleven through twenty, which the authors describe as the half-life of a short-term relationship, the same differences were two to three times larger: d = .45, .39, .22 and .21. For the remaining measures they looked at, the short-term versus long-term differences were consistently small or non-existent throughout.

The authors' own summary is the sentence this page is built on: romantic interest in long-term and short-term relationships began at the same level and rose at the same rate for a substantial period of time — and they specify the unit as months — before the two trajectories diverged. Elsewhere they write that the gradual rise characterising the first several events appeared to be more or less identical, on average, across all relationship subtypes.

That includes the subtypes you would expect to be obvious. One-night stands followed the same rising trajectory of romantic interest across the first nine events before dropping away sharply. Flings and brief affairs were indistinguishable from each other.

So the answer to how many dates it takes is not three, or five, or ten. It is that the early events are where the two kinds of relationship look most alike, and that the signal people are trying to read during them is, on average, the part of the curve that carries the least information. The acquaintance version of the same phenomenon — the finding that agreement about who is desirable decays the longer people know each other, until by around nine months of prior acquaintance it has gone entirely — is set out in what builds attraction over time.

#9

"Short-term" is mostly a label people apply afterwards

Buried in the same paper is a methodological observation that quietly reframes the question, and it deserves its own section.

When the researchers asked, very few participants said they were currently in a short-term relationship: 23% in the first study, and in a follow-up only 8%, against 92% describing one that had ended. The authors offer two readings. Short-term relationships are shorter and therefore harder to catch in progress, which is obviously true. Or the short-term nature of a relationship often becomes clear only with the benefit of hindsight.

The data favour the second. In that follow-up study, the romantic-interest trajectories of ongoing short-term and long-term relationships were nearly identical. The gap only opens up once you are looking at relationships that are over — that is, once the person reporting knows which kind it was.

This is worth sitting with, because it changes what the original question is asking. How many dates before you know presumes that there is a fact about the relationship, sitting there from early on, waiting to be detected by a sufficiently attentive person. What the trajectories suggest is closer to the opposite: for a substantial stretch, there may not yet be a fact of that kind. A short-term relationship is not, mostly, a long-term relationship that failed a test somebody should have administered on date three. It is a relationship that went the same way as the others for months and then stopped.

And this is a recall-based design, so the point cuts both ways and should be stated honestly: people may simply be unable to remember the early doubts they had about someone who later turned out to be wrong. That is a real limit on the finding. It is also, if true, a direct answer to the question — because it would mean the knowing people believe they did in week three is something they constructed later.

#10

Two thirds of relationships did not begin with dates at all

There is a second problem with counting dates, and it is that most relationships do not have a date count to begin with.

Danu Anthony Stinson, Jessica Cameron and Lisa Hoplock looked at how relationships actually start. First they checked what relationship science studies: across the initiation research they located in popular journals and textbooks, only 18% concerned romance that developed between friends. The rest was about strangers and acquaintances — speed dating, first impressions, profiles.

Then they counted. Across seven samples of university students and crowdsourced adults, N = 1,897, the percentage reporting that their relationship began as a friendship ran 40.0, 65.2, 70.7, 72.8, 70.5, 62.2 and 70.6, for a weighted mean of 68.2%. A meta-analysis of the proportions gave a median of 0.66, 95% CI [0.59, 0.72]. The rate did not differ between men and women (67% against 68%), or between students and crowdworkers. It did differ by age — 84% among those under 29, against 65% for people in their forties — and by orientation: 85% among people in same-gender or queer relationships, against 68% for women dating men.

And then the number that matters most for this page. Among friends-first couples, the friendship had lasted a mean of 21.90 months before it became romantic (SD = 28.23), with a median of 12 months and a range from 1 to 180. Only about one in five said they had befriended the person because they were romantically interested to begin with.

For two thirds of people, then, the question is not how many dates it took. There were no dates during the part where the knowing happened; there was a year of being around each other, which is a completely different instrument and a far better one — more contexts, more ordinary Tuesdays, more chances to see how somebody behaves when nothing is being performed. The flip side of that pathway, when it stalls instead of turning, is the subject of why people keep ending up in the friend zone.

One thing this same figure gets used for, it cannot support. Two thirds is a count of how relationships began; nobody in that research was asked how long theirs lasted. Whether the friends-first route produces more durable relationships is a separate question with its own, much thinner evidence base, and we go through all of it in do friends-to-lovers relationships last longer.

#11

What the first few dates are actually good for

None of this means early dates are uninformative. It means they are informative about a narrower set of things than the question assumes, and the narrower set is worth naming.

They are good at ruling out, and poor at ruling in. A first meeting reliably surfaces incompatibility of the factual kind — where someone lives, what they want, whether they are available — and conduct that disqualifies regardless of chemistry. It is much weaker at the positive verdict. The thin-slice numbers say this directly: what reads quickly is sociability, not kindness.

They are the only way to get the information that a profile cannot carry. This is the real content of the speed-dating result, and it is a good argument for meeting sooner rather than exchanging messages for weeks.

They are a poor place to look for intensity as evidence. The relationships that turned out to be short-term started marginally higher on romantic interest, not lower. Strong early feeling is a real experience and it is not a forecast. Where that intensity consists of uncertainty, checking and not knowing where you stand, it is a different thing altogether, and the difference between love and anxious attachment is where to read about it. Knowing your own pattern is worth the two minutes our attachment style test takes, since it changes what early ambiguity feels like.

What does carry information is behaviour over repeated, ordinary contact — how somebody treats people who are not you, what they do when something goes wrong, whether a request is remembered. Those are the markers with observational work behind them, and they are collected in the signs a relationship has real potential. Once there is a relationship to ask about rather than a date count, our relationship satisfaction test reports commitment separately from satisfaction, because in the largest analysis the strongest single predictor was not how satisfied a person was but how committed they believed their partner to be.

And the count can hide a different problem. Waiting for a number to arrive is also what it feels like from inside an arrangement that is not going anywhere in particular, where the absence of a decision is itself the decision; the research on unequal commitment is the page for that. The gap between two people's ages, which the search results attach to this question constantly, is not the thing to count either — in the largest American study to ask, the size of the gap predicted nothing about whether a marriage ended, as what an age gap does and does not predict sets out.

And nothing is bought by waiting on a schedule. The only experiment on timing after a date found that a short wait beat no wait and a long one lost on every measure, which is in the 72-hour rule. Deciding on date three because three is a number is not a decision rule; it is the same folklore with your own calendar attached. More of this under Dating.

#12

What none of this can tell you

The limits go here rather than in a footnote, because on this question they are a large part of the honest answer.

There is no measured number, and this page has not produced one. Two literature searches returned nothing on the question as asked. Anyone offering you three, or five, or ten is offering a hunch, including the hunches that appear under a psychologist's byline.

The trajectory studies are retrospective. People reconstructed relationships they had already lived, which means memory is doing work throughout, and memory is exactly the faculty most likely to smooth an early uncertainty into an early certainty. The authors mitigated this — graphing, event anchoring, within-person comparison, analysing only ended relationships, a preregistered replication — and none of that converts recall into observation.

The samples are narrow. Undergraduates and crowdworkers, overwhelmingly American, overwhelmingly mixed-sex and White in several of the key studies. The friends-first counts come from students and online panels, also retrospective.

The thin-slice work is not about dating. It is judges rating strangers on video, and the condition with the most exposure had the fewest judges in it.

And averages are not people. Every finding here is a statement about what the middle of a distribution does. Some people really do know early and turn out to be right, and nothing in this literature says they are lying — only that, in aggregate, the early period is where the two kinds of relationship resemble each other most, so a confident early read is being made at the point where the evidence is thinnest.

What survives is a change of question rather than an answer to the original one. How many dates until I know assumes the information arrives on a schedule and that the schedule is counted in dates. On the best evidence available, it arrives with accumulated ordinary contact, it arrives in months rather than meetings, and for most people it has historically arrived without any dates at all.

The reverse case — attraction that does not accumulate but collapses in a moment over something small — is covered in what is actually known about the ick, and it rests on a single underpowered study. Every empirical claim on this page carries a linked source, which is the promise made in our editorial policy.

Pulling it together

The question asks for a number, and there is not one. What replaces it is more useful than a number would have been.

The first point is about the instrument. A first meeting reads sociability well and warmth badly; in the thin-slice work, extraversion was judged at r = .42 and agreeableness at r = .11. So the early dates are good at telling you who is good at early dates.

The second is about the curve. In the only study to plot real relationships across their whole span, romantic interest in long-term and short-term relationships began at the same level and rose at the same rate for months before the two separated. The period the question is about is precisely the period where the two kinds of relationship resemble each other most.

The third is the one that reframes it. Very few people describe themselves as currently in a short-term relationship, and the trajectories of ongoing relationships of both kinds are nearly identical. The category is largely applied afterwards, which means there may be no early fact of the matter for an attentive person to detect.

And the fourth is that for about two thirds of people the clock was never counted in dates at all. It was counted in months of ordinary contact, a median of twelve of them, which is a slower instrument and a much better one.

Frequently asked questions

Is it serious after 3 dates?

Nothing in the research makes three a threshold. In the relationship-reconstruction studies by Eastwick and colleagues, which plotted real relationships event by event on a 0 to 100 romantic-interest scale, the relationships that turned out to be long-term and the ones that turned out to be short-term were statistically indistinguishable at the first short date (69.8 against 66.4), the first long date (77.9 against 74.7), the moment of telling the other person you were interested (76.9 against 76.6) and the first kiss (78.2 against 75.0). The differences only became reliable at the sexual events, by which point around twenty events had already happened. Whether something is serious after three dates is a question about what two people have agreed, not about what the third date revealed, and the only way to answer it is to ask the other person.

How many dates before you know, for a man or for a woman?

No study has found a gendered answer to this, because no study has timed the question at all. The closest relevant finding runs against the stereotype and concerns saying so rather than knowing: in the small student samples that asked, men reported thinking about saying "I love you" earlier than women and were more often the first to say it, which is covered in our piece on the signs someone is falling for you. In the friends-first counts, the rate of relationships that began as friendships was 67% for men and 68% for women, a difference that was not significant. Treat the gendered versions of this question as a feature of how the genre is written — most pages on the subject are addressed to one sex — rather than as a finding.

How many dates before becoming exclusive?

There is no measured number, and the research suggests the count is the wrong unit. In the trajectory studies, becoming exclusive is one of 48 events people placed on their own timeline, and the ordering of early events was broadly comparable across relationships that later turned out very differently. What the evidence does support is that this is a conversation rather than a milestone that arrives: perceived partner commitment was the single strongest predictor of relationship quality in a study pooling 43 datasets and 11,196 couples, and what carried the weight was what each person believed about the other's commitment. That belief is produced by being told, not by counting dinners.

Do some people really know immediately?

Some people report it, and what they are reporting has been examined. Zsok, Haucke, de Wit and Barelds measured feelings of love at the moment participants first met potential partners, across an online study, a laboratory study and three dating events. Their conclusion, in their own words, is that experiences of love at first sight were marked neither by high passion, nor by intimacy, nor by commitment, and that physical attraction was highly predictive of reporting it; they read it as a strong initial attraction that some people label as love, either at the time or in retrospect. The full text is paywalled, so no sample sizes are quoted here. The honest summary is that the experience is real and that the label is doing a lot of work.

What is the 7-7-7 rule in dating?

It is one of a family of invented numerical rules — alongside the 3-6-9 rule, the 3-3-3 rule and the three-day rule — that circulate with no study, no author and no origin behind them. Versions of 7-7-7 prescribe a date every seven days, a night away every seven weeks and a trip every seven months. Nothing about the specific intervals has been tested. The nearest relevant evidence is about shared activity rather than scheduling, and it is more specific than the rules suggest: the one randomised field experiment found that activities couples themselves called exciting beat activities they called merely pleasant, while doing more things together did not beat doing nothing. We take the numbered rules apart in our article on the 3-6-9 rule.

How many dates before a kiss?

There is no researched answer, and the question has the same problem as the main one: it converts a mutual decision into a schedule. The trajectory data can say where a first kiss tends to fall in the order of events rather than on a calendar — in the long-term relationships studied it came around the tenth event reported, after flirting, spending time one-on-one and the first dates, and at a reported romantic interest of 78.2 out of 100. It arrived at a very similar point in the relationships that turned out to be short-term. That is a description of what people did, not a recommendation, and nobody is on a timetable.

If I am not sure after several dates, does that mean it is a no?

Not according to the evidence, and this is where the date count misleads most. Agreement about who is desirable decays as people get to know each other: in Eastwick and Hunt's work, consensus slightly exceeded uniqueness among people who had known each other two weeks, while uniqueness exceeded consensus by thirty-seven points among well-acquainted others. Related work found that the longer two people had known each other before dating, the less they matched on attractiveness, until past roughly nine months of prior acquaintance the matching was indistinguishable from zero. And two thirds of relationships in the largest count began as friendships lasting a median of twelve months. Uncertainty at date four is not information about the outcome. It is the normal state of the early part of the curve.

Sources

What this article draws on. For the broader bibliography see our sources page.

Keep reading