# What Actually Decides How a First Date Goes
> Most first date advice has no source behind it. Here is what the research on speed dates, question-asking and nerves actually says decides how one goes.
Written by Ahmet Abiç (Founding editor). Reviewed by Noah Cortez (Stanford Journalism Program). Published August 18, 2026.
Source: https://gritwright.com/social/first-dates/
---
import Figure from '../../../components/Figure.astro';
import ResearchNote from '../../../components/ResearchNote.astro';
import PressNote from '../../../components/PressNote.astro';
import PracticeNote from '../../../components/PracticeNote.astro';
import dateTable from '../../../assets/guides/social/date-table.webp';
import dateListening from '../../../assets/guides/social/date-listening.webp';
import dateWalk from '../../../assets/guides/social/date-walk.webp';
You have an hour to work out whether you want to spend more hours with somebody you have met twice, once on a screen. Both of you know it. That is the whole difficulty, and no amount of preparation removes it.
What preparation can do is stop you running the script that almost everybody runs by default, which is a polite exchange of biographical facts in which neither person says anything a stranger could hold against them. There is recorded evidence of how common that is, and there is measured evidence about the one thing that pulls a conversation out of it.
## The short answer
Most of what circulates as first date advice has nothing behind it. Here is the part that does.
| The question | What the evidence says |
| --- | --- |
| What cannot be known beforehand | Whether the two of you specifically will click |
| What the default date sounds like | Job, hobbies, weekend, both people carefully saving face |
| The one measured lever | Follow-up questions, not new topics |
| What that is worth | Roughly one extra yes across twenty dates, per extra question |
| What to do with nerves | Relabel them as excitement rather than trying to switch them off |
| How to judge it afterwards | By how you behaved and felt, not by whether there was a spark |
## Nobody can screen for this in advance
The largest attempt to predict attraction before two people meet ran [two speed-dating studies](https://pauleastwick.com/s/JoelEastwickFinkel2017PSci.pdf), collected more than a hundred self-report measures from every participant, and fed the lot to a machine learning model. It is worth knowing what the model could and could not do, because the difference is the entire argument for going on the date at all.
Unattached participants filled in **more than 100 measures** of traits and preferences that previous research had linked to mate selection, then went on a **four-minute date with every opposite-sex person at the event**.
Random forests models predicted **4% to 18% of actor variance**, meaning how much a given person tends to desire others in general, and **7% to 27% of partner variance**, meaning how much others tend to desire that person.
On relationship variance, the part that captures whether these two people in particular want each other, the models **could not predict it using any combination of traits and preferences reported before the dates.**
The authors' conclusion: compatibility elements of human mating are challenging to predict before two people meet.
Read that carefully, because it cuts two ways.
Some of you is predictable. How generally enthusiastic you are about other people, and how much other people tend to warm to you, are stable enough to be partly modelled from a questionnaire. That is the actor and partner variance, and it is the part that responds to the ordinary things: being interested in people, being pleasant to be near, having a life you are not embarrassed by.
The specific thing, though, the reason this particular evening either works or does not, was invisible to a model holding a hundred measurements on both people. Paul Eastwick's own comparison was that romantic desire may be more like an earthquake, a dynamic and chaos-like process, than a chemical reaction involving the right combination of traits.
So the frame most people bring, that a first date is an examination they might fail if they prepare badly, is wrong on the facts. There is no preparation that determines the outcome, because the outcome was not determined by anything either of you knew about yourselves an hour earlier. What you can influence is narrower and more useful: whether the hour gives the thing a chance to happen at all.
This is the same finding that runs through [what actually makes a relationship work](/social/what-makes-a-relationship-work/), where a machine learning model across 43 longitudinal studies found the same shape. What each person brings predicts a lot. What the pairing produces resists prediction.
## What a first date actually sounds like
The gap between how people imagine first conversations and how they go is documented, and the documentation is German. A linguist at the University of Münster [recorded and transcribed 134 real speed dates](https://www.uni-muenster.de/news/view.php?cmdid=9021), then analysed them as conversations rather than as outcomes.
Franz analysed **134 recorded speed-dating conversations** for her doctoral work in German linguistics.
Participants turned out to be **far more reserved and conservative than expected.** Conversations ran through a predictable set of standard questions about work and hobbies, with very little flirting. The university's own headline for the finding translates as: at speed dating, there is rarely a spark.
Participants also played down why they had come, saying they were there just for fun, or that Sunday is a dead day anyway. And they avoided passing any judgment on the other person, positive or negative, in what Franz describes as face-saving.
That last detail is the one worth sitting with. It is not that people were boring. It is that both people were actively working to say nothing that could be held against them, and the safest way to do that is to trade CV facts and withhold reactions.
Which produces a conversation that is pleasant, correct, and gives neither of you a single piece of information you could not have read on a profile. Then you go home and try to decide whether you liked somebody you never met, because what you met was their public-facing summary.
The interview feeling that everyone complains about is not caused by nerves. It is caused by two people simultaneously deciding that safety is the priority.
## The one behaviour that moves the number
There is exactly one conversational behaviour I could find with [a measured effect on whether the other person agrees to a second date](https://www.hbs.edu/ris/Publication%20Files/Huang%20et%20al%202017_6945bc5e-3b3e-4c0a-addd-254c9e603c60.pdf), and it is not a topic, an opener, or a piece of body language.
Across three studies of live conversations, **people who asked more questions, particularly follow-up questions, were better liked** by the person they were talking to.
The researchers hand-coded question types across 368 transcripts and found six kinds. **Follow-up questions were the most common**, at around 40% of all questions asked, ahead of full-switch questions that change the subject.
They then trained a follow-up question detector and ran it over recorded speed dates. Speed daters who asked more follow-up questions were **more likely to get agreement to a second date**, which is a behavioural measure of liking rather than a self-report.
Their explanation is responsiveness: a follow-up question is evidence that you were listening, understood, and cared enough to want more.
The size of the effect is the part that makes it worth changing your behaviour over, and [Harvard's own write-up of the study](https://www.library.hbs.edu/working-knowledge/asking-questions-can-get-you-a-better-job-or-a-second-date) puts a number on it.
The speed-dating analysis covered **110 speed daters across more than 2,000 conversations**, drawn from an earlier Stanford dataset.
**The top third of question askers got the most second dates.** The researchers estimated that a participant asking **just one more question on each of their 20 dates would get one additional "yes, I want to see you again"** on average.
The same piece carries the warning that keeps this from being a trick: **asking a barrage of questions without disclosing anything yourself can read as guarded, or as invasive.**
One extra question per date, one extra yes in twenty. That is a small effect honestly reported, and it is more than any other single thing in this article can claim.
It also explains itself. A follow-up question is the only conversational move that cannot be faked in advance, because it has to be built out of what the other person just said. Prepared questions signal preparation. Follow-ups signal attention, and attention is the scarce thing at a table where both people are rehearsing.
## How to ask follow-ups without interrogating anyone
The failure mode here is obvious and the researchers named it: a person who only asks questions comes across as guarded or as conducting an inquiry. The fix is not fewer questions but a different rhythm.
**Answer before you ask.** When they ask you something, give a real answer with something in it, then follow up on theirs. The disclosure is what stops the questions reading as an interview, and it is the half most people drop when they are nervous.
**Follow the detail, not the headline.** They mention they moved cities last year. The headline question is where from. The follow-up is what made you do it, or what you miss. One collects a fact, the other opens a subject.
**Ask about the thing they got animated about.** People signal interest before they say it. If their voice changed on a subject, that is the thread, whether or not the subject interests you yet.
**Do not stack.** Two questions in a row without a response in between turns into a form. Ask one, let the answer land, react to it before you ask the next.
**Let a silence sit for a second longer than is comfortable.** Most people fill a pause with a new topic, which is a full switch, which is the weaker move. Often the other person was about to say the more interesting half.
None of this requires you to be quick or charming. It requires you to be listening closely enough that you have something specific to ask, which is a different skill and a learnable one.
## Going deeper, on purpose
If the default conversation is safe biography and the improvement is following what the other person says, the obvious next question is how far you can push that in an hour. There is [a body of laboratory work on exactly this](https://psychology.uga.edu/sites/default/files/RCITarticle1999.pdf), because researchers needed a way to make two strangers feel close on demand.
The Relationship Closeness Induction Task is a structured self-disclosure procedure: **29 questions in three lists, administered over nine minutes**, one minute on the first seven questions, three on the next twelve, five on the last ten.
The lists get **progressively more personal**, which is the active ingredient. The authors build on the principle that a vital feature in the development of a close relationship is **reciprocal and escalating self-disclosure.**
It has been used successfully in five experiments. Participants report **feeling closer, feeling more similar, liking each other better and being more likely to be friends in future** than a control group, with alphas from .78 to .89, and **no gender differences** have been found.
Nine minutes. That is the finding that should change how you think about an hour in a bar.
The better known version of this is the 36 questions, from a 1997 study by Arthur Aron and colleagues, which uses the same escalating structure over 45 minutes. That paper sits behind a publisher paywall, so rather than cite something I could not open, here is the procedure as [Berkeley's Greater Good in Action reproduces it](https://ggia.berkeley.edu/practice/36_questions_for_increasing_closeness), with all 36 questions and the three-set timing.
Now the honest part. You are not going to pull out a numbered list on a date, and you should not. Both of these are laboratory procedures, run with instructions and a timer, and the effect is on felt closeness in that moment rather than on anything durable.
What transfers is the principle underneath, which is the same one the follow-up question research found from the other direction. Closeness comes from reciprocal escalation. Both people, going slightly further in, taking turns. Not one person confessing, which is unnerving, and not both people staying at the level of what they do for work, which is what 134 recorded speed dates turned out to be.
In practice, that means being the one who goes one step past the safe answer, then leaving room. Ask what they are actually working towards rather than what they do. Say the true version of why you left your last job instead of the diplomatic one. When they ask what you do at weekends, answer with something you care about rather than something defensible.
The risk is real and small: you say something slightly too honest and it lands badly. The alternative is the interview, and the interview has a ceiling.
## What to do about the nerves
The standard instruction is to calm down, and there is [a study showing that almost everybody believes this is the right approach](https://www.hbs.edu/faculty/Pages/item.aspx?num=45869) and that it is the harder route.
Brooks found that **an overwhelming majority of people believe trying to calm down is the best way to cope** with anticipatory anxiety.
Across studies involving **karaoke singing, public speaking and maths performance**, she tested the alternative: reappraising anxiety as excitement. Compared with those who tried to calm down, people who reappraised **felt more excited and performed better.**
The intervention is minimal. **Saying "I am excited" out loud**, or simply being told to get excited, was enough to shift people into an opportunity mindset rather than a threat mindset.
Her explanation is arousal congruency: anxiety and excitement are both high-arousal states, so relabelling is a short move, while calming down is a long one.
No date was tested in that paper, and I am not going to pretend otherwise. What was tested is the general case of anticipatory nerves before a performance you care about, which is the same machinery.
The practical form is unglamorous. Your hands are unsteady and your pulse is up on the way there. That state is not evidence that something is wrong, and it is not going to switch off in the next ten minutes because you told it to. It is the body preparing for something that matters, which is accurate, because it does matter, a bit. Naming it as anticipation rather than dread costs nothing and has a measured effect on performance in the studies where it was tested.
The thing that genuinely lowers the stakes is arithmetic rather than psychology. If a first date is an examination with one attempt, the nerves are proportionate. If it is one of many evenings you will spend meeting people, most of which will be pleasant and inconclusive, they are not. The second description is the true one.
## Where you go changes what can be said
Venue is treated as a matter of taste and it is closer to a matter of structure, because the format determines which conversations are available to you.
**Sitting across a table** is the default and it is the most exposed arrangement there is. Constant eye contact, nothing to do with your hands, and every silence fully audible. It rewards people who are fluent and punishes people who are not, which is why it produces so many interviews.
**Walking** removes the eye contact requirement and gives silences somewhere to go. This is not a research finding, it is a structural one, and it is the same reason difficult conversations in [long relationships](/social/how-to-stop-fighting/) tend to happen in cars.
**An activity** solves the silence problem and creates a different one: it can fill the entire evening, and you can leave having learned that they are competent at bowling. Useful as half of a date, risky as all of it.
**Anything loud** removes the possibility of the conversation this whole article is about. If you cannot hear a follow-up answer, you cannot ask a follow-up question.
The reliable shape is a short first thing with a natural ending, somewhere you can hear each other, with an obvious way to extend it if you both want to. Which brings us to the part nobody plans.
## How to end it
Most first dates end badly in a specific and avoidable way: they run past the point where either person is enjoying it, because leaving feels like a verdict.
Set the end before you arrive. Tell them you have something at nine, or just plan a first thing that finishes on its own. This is not a trick to seem busy. It removes the moment where you both have to decide, in real time, in front of each other, whether this is going well enough to continue, which is a decision neither of you can make reliably at that point anyway.
Then, if you want to see them again, say so out loud before you separate, and be specific enough to be actionable. There is no strategic reason to withhold it. The research on second dates is about the person deciding, not about how cool you looked while waiting.
If you do not want to, do not say you will. The polite lie is easier for ten seconds and worse for everyone afterwards, and this is the same territory as [saying no without starting a fight](/social/how-to-say-no/), which is a skill worth having early.
## Judging it afterwards, without asking about the spark
The question everybody asks in the taxi is whether there was a spark. It is a bad instrument, for the reason [the machine learning study](https://pauleastwick.com/s/JoelEastwickFinkel2017PSci.pdf) demonstrated: the thing you are trying to detect was not detectable an hour before, and it is not reliably detectable an hour after either.
Ury built a checklist she calls the post-date eight in 2019, for a coaching client who kept dismissing matches for failing to meet a list of surface criteria. She published it, and has taught it since.
The eight are all about you rather than about them: **what side of me did they bring out; how did my body feel; do I feel more energised or less than before; is there something about them I am curious about; did they make me laugh; did I feel heard; did I feel attractive in their presence; was I captivated, bored, or somewhere between.**
Her instruction is to run through them on the way home, and her claim is that this trains attention away from instant chemistry and towards whether interest is trending upwards.
This is practice rather than a study, and it is presented here as exactly that. Nobody has run a controlled trial on the post-date eight, and Ury's evidence is her own coaching practice.
The reason it earns a place is that it asks a question you can actually answer. Whether you clicked with somebody is a judgment about a relationship that does not exist yet. Whether you were more yourself or less, more energised or less, curious or not, are observations about the last two hours, and you were there for those.
The practical version is even shorter. **Do you want to see them again? Are you curious about anything specific?** If you cannot answer either, that is information too, and it usually means go once more, because a second date resolves it faster than a week of thinking.
## What this does not fix
Everything above is about the hour itself, and the hour is not where most dating goes wrong.
It does not help you meet anyone. That is a separate problem with separate answers, and [meeting people as an adult](/social/how-to-make-friends-as-an-adult/) covers the general version of it.
It does not make an unpromising match promising. Follow-up questions produce a better conversation with somebody you were never going to want, which is a nicer evening and the same outcome.
It does not tell you whether to continue past the second or third date, which is a harder question than the first one and gets its own guide.
And it will not stop most of them going nowhere. The base rate is what it is. The finding at the top of this article, that the specific chemistry between two people resisted prediction from a hundred measurements, is also the reason the process involves volume. That is not a failure of technique. It is the shape of the thing.
What you can control is that the hour was spent on an actual conversation rather than a mutual recital of biography, so that whatever was or was not there had a chance to show up. That is the whole available edge, and it is worth having.