First Date Mistakes: What the Research Says Actually Costs You

What this page used to say, and why it changed

This article was a list of eleven mistakes, and most of it was written as personal experience: an author who had made seven of the ten mistakes, had worn the wrong dress, had learned about loud bars the hard way. None of that was anyone's experience. It was a device, and a publication that promises to show its evidence cannot build an article on anecdotes that did not happen. The first person is gone.

Four claims went with it. The list cited Mehrabian's 1971 book, the source of the famous 7-38-55 split, which is about how people judge feelings from inconsistent tone and words and has never supported the idea that most communication is non-verbal. It credited Gottman's research with a finding about "presence" on first dates that is not in any paper of his we could locate. It advised talking less, which turns out to be the opposite of what the only experiment on the question found. And its closing line told readers to keep the phone in a pocket — a piece of advice the same article's own phone section had already retracted.

What replaces them is organised by how much evidence each point has, which is not how first-date lists are usually organised. Some of the advice below rests on experiments with thousands of people. Some rests on a single study of forty speed dates. And several of the mistakes everybody warns about — the ex, the outfit, the loud bar — have never been tested at all, which is said where it applies. The full record of what changed is on our corrections page.

The mistake with the most evidence behind it: not asking the second question

If one finding on this page is worth remembering, it is this one, because it has an experiment and a field study pointing the same way.

Karen Huang, Michael Yeomans, Alison Wood Brooks, Julia Minson and Francesca Gino, in the Journal of Personality and Social Psychology in 2017, first ran it as an experiment. 398 people in 199 pairs chatted online, and one member of each pair was told to ask either many questions or few. The partners of the people told to ask many liked them more (5.79 against 5.31 on the scale, d = .35), and a second experiment with 169 more pairs found the same thing (d = .27).

Then they went to speed dating. 110 daters across three sessions each had fifteen to nineteen four-minute dates, 1,961 date observations in all, transcribed and coded by question type. The average dater asked almost ten questions per date. What predicted getting a yes for a second date was not the number of questions. It was the rate of follow-up questions — the ones that pick up what the other person just said. In the authors' own illustration, a difference of about 8% in the follow-up rate went with one additional second date over an evening of twenty. Other kinds of question — introductions, switches of topic, mirrored questions — did not predict success.

Three caveats, stated because the popular version of this study drops all of them. The speed-dating result is correlational, and it weakens once you account for how desirable each asker already was. It was challenged in a reanalysis by Avraham Kluger and Thomas Malloy in 2019, to which the authors replied in the same journal with a further analysis of follow-up questions. And in 2025 the paper was independently audited and corrected: the correction reports several minor errors and states that the conclusions remain valid.

The mechanism the authors propose is the useful part. A follow-up question is evidence of listening, and listening is read as responsiveness — understanding, validation, care. That is also why a list of prepared questions misses the point: the questions that worked were the ones that could not have been prepared, because they depended on the previous answer. For why the feeling of being understood matters more than the information exchanged, see what the research says about how people open up.

The advice that is probably backwards: talk less

Almost every first-date list, including the old version of this one, tells nervous people to stop talking so much. There is exactly one experiment that tested it, and it points the other way.

Quentin Hirschi, Timothy Wilson and Daniel Gilbert published it in Personality and Social Psychology Bulletin in 2022 under the title "Speak Up!". First they asked people what they would do. In two forecasting studies (95 and 110 participants), people who wanted to be liked said they would speak for less than half of a conversation — 43% and 45% on average. People who wanted to seem interesting said they would talk more, about 58%.

Then they tested it. 116 undergraduates were paired with strangers and a computer assigned each person a share of a conversation of just under seven minutes: 30%, 40%, 50%, 60% or 70%. The result was a linear trend: the more people spoke, the more they were liked. Speaking time made no difference to how interesting people were judged to be, and none to how much their partners enjoyed the conversation. People were, in other words, planning to talk less than would have served them, for a goal — being liked — that talking less did not help.

This is a small study with a barely significant result (p = .045), all the participants were American students, and the talking was assigned rather than chosen. The authors say themselves that someone who chooses to hold the floor for 70% of the time may well be read differently. So the finding does not license a monologue. It does knock out the specific fear behind "talk less": that each extra minute of you costs something.

The better-supported version is about the shape of the exchange, not the share. Susan Sprecher and colleagues had 78 pairs of strangers either take turns disclosing or split the job, one person disclosing while the other listened and then swapping. Pairs who took turns reported more liking, closeness, perceived similarity and enjoyment, and the difference persisted after the second session, when the other pairs had had their turn too. Long single turns produced less liking than quick back-and-forth. The mistake, then, is not talking. It is talking in blocks that do not leave the other person a way in — which is what the follow-up questions in the section above prevent.

Disclosure: the evidence is about timing, not about how deep you go

A second piece of standard advice is to keep a first date light. The literature on self-disclosure mostly does not support that either, but it does support something narrower.

Nancy Collins and Lynn Carol Miller's 1994 meta-analysis in Psychological Bulletin found three separate effects: people who disclose are liked more (94 effect sizes), people disclose more to those they already like, and people come to like those to whom they have disclosed. Measuring disclosure by its intimacy gave larger effects than measuring its amount. The authors expected very intimate disclosure to backfire, and found little evidence that it did, though very few studies tested it.

What the experiments do show is that when a disclosure arrives matters. Camille Wortman and colleagues, in 1976, had a confederate make the same intimate disclosure either early in a ten-minute conversation or near the end. The early discloser was liked significantly less and described in more negative terms. We have read the abstract of this study, not its full text, so its sample size is not reported here.

The ex is where this lands, because "don't talk about your ex" is the most repeated first-date rule there is. We looked for a study that tested it — mentioning a previous partner on a first date, measured against attraction — and did not find one. What can be said is the general version: a heavy disclosure early reads worse than the same disclosure later, and an account of a recent breakup is usually a heavy one. A passing, factual mention is not the same thing, and nothing shows it costs anything. If the ex keeps arriving whether invited or not, the useful question is not how to hide it but whether you have actually moved on, which the breakup research answers with something other than a timeline.

The one experiment that gets cited for everything in this area is Arthur Aron and colleagues' 1997 procedure — the thirty-six questions. It is worth knowing what it did and did not show: pairs of strangers who spent 45 minutes on questions that got gradually more personal ended up closer than pairs given small-talk questions (4.06 against 3.25 on the closeness scale), but the average closeness they reached sat at roughly the 30th percentile of how close students felt in their closest relationship. It builds closeness between strangers. It does not produce love, and the couple who famously married came from an earlier version the paper itself calls anecdotal. Whether any set of questions can tell you about compatibility is a separate question with its own evidence.

Trying to make a good impression is not the mistake it is made out to be

"Just be yourself" is the other universal instruction, usually paired with the claim that trying to impress someone backfires because it makes you stiff and exhausted. Both halves are shakier than they sound.

The best direct test is small. Lauren Human, Jeremy Biesanz, Kate Parisotto and Elizabeth Dunn, in 2012, recorded 24 people answering getting-acquainted questions on video. Eleven were told to put their best face forward — in the paper's words, as they would with someone they had just met or just started dating — and thirteen were told to minimise any effort to make an impression. 66 viewers then rated all of them. The people trying to present well were liked more, and — the surprising part — seen more accurately: viewers' impressions of their personalities matched the targets' self-reports better. The link ran through engagement. Trying to come across well made people more animated, and animation gave the viewers more to go on. Two limits matter. It is a sample of 24, and the instruction explicitly told people not to play a role: it compared positive-but-genuine effort with low effort, not honesty with pretence.

The "exhausting" half comes from the ego-depletion literature. Kathleen Vohs, Roy Baumeister and Natalie Ciarocco reported in 2005, across eight experiments, that self-presentation which went against a person's habits drained their self-control afterwards, and that already-drained people presented themselves worse — including talking too much and disclosing too much or too little. Every one of those experiments rests on the depletion paradigm, and that paradigm has not survived replication: a preregistered test across 23 labs and 2,141 people found d = 0.04, and a second one of 36 labs and 3,531 people, led by Vohs herself, found d = 0.06 and no confirmatory effect. The idea that impression management uses up a limited resource should not be presented as established.

What does hold up is the ordinary version: presenting a version of yourself you cannot keep up only postpones the moment the difference shows. That is a point about honesty, not a finding about attraction, and the research above suggests that effort in the direction of who you already are is not the problem. The same caution applies to the body-language shortcuts — the expansive "power pose" that was supposed to change how confident you are failed its large replication, which is covered in what body language does and does not reveal.

Your nerves are much less visible than they feel

Most first-date mistakes, looked at closely, are anxiety wearing a disguise: over-preparing, over-talking, filling every silence, reading every pause as a verdict. There is one well-tested finding that speaks directly to the anxiety itself.

Kenneth Savitsky and Thomas Gilovich called it the illusion of transparency. In their 2003 study in the Journal of Experimental Social Psychology, people giving an unprepared talk rated themselves as looking considerably more nervous than their audience rated them: the inner state felt obvious from the inside and mostly was not visible from the outside. In a second study, simply telling speakers about the illusion before they spoke improved their performance, as judged both by themselves and by observers. Nothing in it is about dating specifically, and the sample sizes are in the full paper, which we have not reproduced here.

The practical point is modest and real. The nervousness that makes people over-talk, or rush the ending, is mostly a problem inside the person having it. The date across the table is seeing a fraction of it — and the table is not the place the nerves do their most damage. That comes afterwards, in the reading of how it went.

The mistake after the date: deciding it went badly

The most common first-date error may happen on the way home.

Erica Boothby, Gus Cooney, Gillian Sandstrom and Margaret Clark named it the liking gap in Psychological Science in 2018. Across five studies and roughly 440 people — strangers in a lab, people at workshops, first-year students who had just been assigned to share a room — people consistently underestimated how much their conversation partner had liked them and enjoyed talking to them. Observers who watched recordings could see the partners' liking; the people in the conversation could not, because they were busy reviewing their own performance. The gap was larger for shyer people, paying for accurate guesses did not remove it, and among the roommates it lasted several months before disappearing near the end of the year.

Two practical consequences. The first is that your own sense of how a date went is biased downwards, so "it was fine but I probably came across badly" is a weak reason not to follow up. The second runs the other way and is covered in the section on endings below: the other person's politeness is also not a reliable reading of what they want. The fix for both is the same, which is to ask rather than infer. When to send that message is a researched question with a specific answer, and it is in the 72-hour rule. For people whose anxious reading of a date goes on for days, the attachment piece on anxious love is the better next step.

The venue: the thrill-bridge advice rests on less than it looks

Some first-date advice recommends the opposite of a quiet dinner: go somewhere exciting, because arousal gets misread as attraction. It traces back to one of the most famous studies in social psychology.

In Donald Dutton and Arthur Aron's 1974 experiment, a female interviewer approached men on either the swaying Capilano Suspension Bridge in Vancouver or a low, solid bridge, and gave them her number in case they had questions about the study. 9 of 18 men from the frightening bridge called, against 2 of 16 from the safe one. The cells are tiny, one interviewer carried the whole effect in that experiment, a male interviewer produced no difference at all, and nobody measured the arousal of the men who were actually tested.

The wider literature is more careful. A 1998 meta-analysis by Craig Foster and colleagues pooled 33 experiments and found a real but small effect of arousal on romantic attraction (d = 0.27) — with a catch that makes it useless as dating advice: arousal polarised judgements. It increased attraction to people who were already attractive to the rater and decreased it towards those who were not. It amplifies what is there. It does not create it.

And the one field test that looked at the person you are actually with found nothing for them. Cindy Meston and Penny Frohlich surveyed 251 people at the entrance and exit of roller coasters in Texas. People riding with a friend or relative rated a photo of a stranger as more attractive after the ride. People riding with a romantic partner did not, and ratings of the partner themselves did not rise either. The sample was not randomised, and partner ratings were so high to begin with that there was little room to go up.

So an exciting venue is fine if both of you would enjoy it, and nothing shows it manufactures attraction to you. The case for somewhere you can actually hear each other is not a study. It follows from the sections above: the behaviours with evidence behind them — follow-up questions, turn-taking — all require a conversation, and a room that makes one impossible removes them.

What to wear: the red-dress finding shrank

There is no research on overdressing or underdressing for a first date. There is research on one clothing claim, and its trajectory is a good example of how this literature works.

Andrew Elliot and Daniela Niesta reported in 2008, across five small experiments, that men rated women as more attractive and more sexually desirable when the women were pictured in red or against a red background. The cells were small — the first experiment had 27 men — and the effect was large.

A decade later, a meta-analysis by Gabrielle Lehmann, Elliot himself and Robert Calin-Jageman pooled the replications. For men rating women: 45 studies, 2,961 people, d = 0.26, varying a great deal from study to study and shrinking over time. For women rating men, smaller still and with signs of publication bias. In the preregistered studies the effects were close to zero. The authors did not agree on what this means: two of the three concluded that the true effect is very small and possibly nonexistent; Elliot's reading was that a small effect exists for male observers and needs better methods.

That is the whole evidence base for dressing strategically on a date. Wear what fits the place and what you are comfortable in; there is no study showing more than that, and the one colour finding everyone quotes has mostly faded under replication.

Warmth that goes to everyone reads as nothing in particular

One piece of first-date advice that does have evidence behind it concerns selectivity, and it is usually told the wrong way round.

Paul Eastwick, Eli Finkel, Daniel Mochon and Dan Ariely studied 156 speed daters in 2007. People who were especially attracted to one particular partner — more than to their other partners — tended to have that attraction returned, a modest effect (r = .14). But people who said yes to everybody were, on average, desired less by the people they met (r = −.41), and their partners reported less chemistry with them. Being interested in someone in particular was attractive. Being interested in everyone was not.

It is a short report from one sample, and the explanation the authors tested — that indiscriminate people were seen as unselective — was only marginally supported. But it undoes the idea that more warmth, spread more widely, is always better. The specific version of that advice, about flirting with the server or the bartender, has never been studied; the general version has, and it says interest registers as interest when it is specific. The mirror problem — mistaking general friendliness for interest in you — is covered in flirting or just friendly, where it turns out the common error is the opposite one.

Reaching for the phone mid-conversation — the table is not the mistake

An earlier version of this page said a phone visible on the table was the one item on the list with real data behind it. That was exactly backwards.

The mere presence of a phone has been tested six times and pooled: an average correlation of −.20 across 601 participants, not statistically significant. The original 2013 experiments by Andrew Przybylski and Netta Weinstein, with 37 and 34 pairs, were directly replicated by Cristina Linares and Anne-Laure Sellier in 2021 with 356 people, and the effect on relationship quality did not appear. Put the phone wherever you like.

What is grounded is picking it up while the other person is talking. That is a different behaviour, with 37 studies and 7,515 participants behind it, and the mechanism that has been tested is not rudeness but exclusion: being snubbed for a phone threatens the sense of belonging roughly the way social exclusion does. A check that visibly has nothing to do with your date still costs something, which is why saying what it is for costs a sentence and buys most of it back.

The full account, including where these numbers are weak and why people's estimates of their own phone use are close to worthless, is in what phone use on a date actually shows.

Ending the date: a polite yes is not always a yes

The last minutes of a first date are where the most avoidable damage happens, and the reason is a finding about how hard it is to say no to someone in front of you.

Samantha Joel, Rimma Teper and Geoff MacDonald had single students choose the dating profile they liked best and then learn something off-putting about that person — an unattractive photo in one study, the student's own stated dealbreakers in the other. Some were told the person was real and in the building; others only imagined it. Those who believed it was real were much less willing to reject — in the first study, going by the authors' posted data, roughly 16% of those imagining the choice agreed to exchange contact details, against about 37% of those who thought it was real. The gap was partly explained by concern for the other person's feelings, which people who were only imagining it did not anticipate. The same finding, seen from the side of the person hoping for a yes, is at the centre of why some people keep ending up in the friend zone.

That has an uncomfortable implication for the end of a date. A yes to a second date, a hug, or a kiss, given face to face, is partly shaped by not wanting to hurt you. Asking in a way that is easy to decline is not a nicety. It is the only way to get an answer that carries information. Something like: "I had a good time — I'd like to do this again, if you would too," and then leaving room for a no.

A second finding points the same way. Adam Mastroianni and colleagues asked people in 932 conversations when they had wanted them to end. Conversations almost never ended when both people wanted them to, and people had very little idea when the other person had wanted to stop. So "it felt like they didn't want it to end" is not a reliable reading either. Nobody has studied the ideal length of a first date, and any number you read — ninety minutes, two hours — is a guess.

Where this stops being a matter of manners. Research on how students communicate sexual consent — Kristen Jozkowski and colleagues surveyed 185 of them — found that men relied more on non-verbal cues to read a partner's consent, and women reported more verbal ways of giving or refusing it: the two sides were reading different channels. That is the case for asking in words. And continuing to press after a no, or after a hesitation, is not a first-date mistake. It is pressure, and past a point it is coercion, whatever the evening was like until then. If a date has moved from awkward to frightening, the support lines are listed in the sources below.

Chemistry is not a compatibility reading — and neither is your checklist

The last mistake is the slowest one: treating a first date as a verdict on the relationship, in either direction.

The strongest evidence on this is humbling for everybody involved. Samantha Joel, Paul Eastwick and Eli Finkel took 350 speed daters who had each filled in more than a hundred measures of their traits and preferences in advance, and asked machine-learning models to predict who would desire whom. The models could predict, somewhat, how much a person desired others in general and how much they were desired in general. They could not predict the part that is specific to a pair — the unique attraction of one person to another, which was the largest share of the variance — at all. Whatever makes two particular people click was not in anything either of them could report beforehand. That result is also the reason a dating app's matching has never been shown to work, since a sign-up questionnaire is exactly the kind of data the models here were given.

Stated preferences fare no better. In Eastwick and Finkel's 2008 speed-dating study of 163 students, the ideals people wrote down in advance — including the familiar sex differences in how much looks and earning prospects mattered — failed to predict whom they actually desired at the event, where those sex differences disappeared. And a meta-analysis by Matthew Montoya, Robert Horton and Jeffrey Kirchner, covering 313 studies, found that actual similarity predicted attraction when people had not met or had only just met, but was not a significant predictor in existing relationships; perceived similarity mattered throughout. The checklist you arrive with is a weak guide to what will matter.

There is one measurable thing that does track mutual interest, and it cannot be faked on purpose. Molly Ireland and colleagues analysed the language of 40 four-minute speed dates: pairs whose styles of speech converged — function words, the small connective tissue of sentences — were about three times more likely to want to see each other again for each standard-deviation increase in matching. It is one small study with a borderline result, and matching is a sign of mutual engagement rather than a technique. But it fits the rest of this page: the signals that carry information are the ones produced by paying attention.

So a flat first date is weak evidence against a person, and a brilliant one is weak evidence for them. Agreement between observers about who is desirable falls the longer people know each other, which is where what builds attraction over time starts, and what to watch once a few dates have happened is covered in the signs a new relationship has real potential. Chemistry read as compatibility is also how people end up months into something with no shape, which is what an undefined arrangement is actually evidence of. If the same first date keeps going the same way with different people, the pattern behind repeatedly choosing the wrong people and the attachment style test are more useful next steps than another checklist. More first-date material is under Dating.