How to Make Small Talk When You're Bad at It

Learning how to make small talk starts with a single switch: describe, do not retrieve. The freeze happens because you are searching memory for something interesting, under time pressure, while watching your own face. That is a lot to run in three seconds. Describing what is already in front of both of you costs almost nothing and works just as well.

Key Takeaways:

• The blank mind is a search problem, not a personality problem. Retrieval under time pressure fails; description of the shared situation does not
• Decide who you will speak to and what you will say before you enter the room. The decision, not the sentence, is what eats the window
• Open with a statement about the situation rather than a question about the person, because a statement costs the other person nothing to ignore
• Both people in a first conversation walk away with the other person’s opinion of them set too low; that is the liking gap, and it makes your own verdict useless as evidence (Boothby et al., 2018)
• The practice protocol below targets the dread on the approach rather than the fluency, because the dread is what keeps people out of the room

Why You Freeze, and What That Is Not

The freeze is a search failure. Standing at the coffee machine, you are trying to retrieve an interesting item from memory, judge whether it is appropriate, and deliver it inside about three seconds. Nobody performs that well. The people who look natural at small talk are not searching, because they are describing what is already in the room, and description requires no retrieval at all.

Two processes compete for the same attention: producing speech and grading it. Grading is the expensive one. It is also the one that produces the blank, because attention spent auditing the last sentence is attention not available for the next one, and the loop tightens every time you notice it tightening.

None of this is shyness in the sense people mean it. Shy people freeze and so do loud people, and plenty of people who present well in meetings go blank in a break room. If avoidance has started shrinking your life, that is a separate and treatable thing worth taking to a professional. Everything below addresses the mechanical part. It is most of it, for most people.

Stop treating the blank as evidence about you. Treat it as evidence that you asked your memory for the wrong thing.

The Sequence: How to Make Small Talk in Four Moves

The four moves run in this order, and the choosing happens before you are standing there.

1. Choose before you arrive. Walking into a room, pick one person and one opener while you are still at the door. Look for someone alone, someone at the food, someone with a phone face down. The point is that the choosing happens when you have 30 seconds to do it, not when you have three, because the deciding is what actually eats the window.

2. Say a statement about the situation. Not a question about them. “That line has a personality and it isn’t good,” “these chairs are a punishment,” “I’ve never seen this machine work on a Monday.” A statement gives three exits instead of one: agree, correct you, or add something. It also costs the other person nothing to ignore, which is exactly why it is safe to make. The situation you are both standing in is the most reliable of all small talk topics: it needs no lead-up.

3. Do one thing with the answer. One thing, not a plan. Their reply contains a noun you did not put there, so ask about that noun. They say they are in early for a call, and you ask whether the call was worth it. Question-asking experiments from Huang and colleagues in 2017 found the liking effect concentrated in follow-ups rather than spread across questions in general, which is a relief for anybody reading this article: the highest-value move is the one that requires no material of your own. The follow-up is where small talk questions actually earn their keep.

4. Add one thing of your own, the same size. After they hand you something, hand back something similar in weight. They mention a bad week, you mention yours in one sentence. Matching is what converts an interview into a conversation, and skipping it is why some people get answers all evening and never get a friend.

When you want out, say a reason and a closing line without waiting for a gap.

If the four moves feel abstract, a set of annotated small talk examples shows them landing in real exchanges. Run those four in a supermarket line this week. The whole sequence takes 40 seconds and nothing about it requires you to have thought of anything clever.

What to Do When It Stalls Mid-Sentence

Stalls have three shapes, and each has a specific repair rather than a general one.

The blank. You have opened and now nothing is coming. Say what you see. “I have completely lost my train of thought” also works, and it works better than people expect, because it is honest, brief, and hands the other person an easy turn.

The dead answer. You asked, they gave you one word, and the silence is now yours to fill. Do not ask a second question immediately, since two questions in a row after a one-word answer reads as an interrogation. Say something about yourself instead, one sentence, and let them decide whether to pick it up.

The person who does not want to talk. Feet toward the door, eyes on the phone, three-word replies, the small half-turn back toward the counter. Let it go on the first signal. This is the repair beginners skip: they read a normal decline as a personal failure and then either push harder or replay it for two days. Whatever is making that person decline was in place before you opened your mouth.

One caution about the repairs. They work when used once. Announcing your own awkwardness twice in the same conversation turns it into the topic, and then the other person has a job.

Whichever of the three you hit most is the one to practice, and it is almost never the opener.

What Does Not Work

Rehearsing lines in the car does not work, and it is the most common preparation people do. Memorized lines come out with the wrong timing, and worse, they set up the very retrieval task you are trying to avoid: now you are searching for the prepared sentence instead of looking at the room.

Trying to be interesting does not work either. It raises the bar for every sentence, which slows production and adds monitoring, and it aims at the wrong target. Nobody at a coffee machine is evaluating you for interestingness. They are deciding whether talking to you is easy.

Waiting for a natural opening does not work, because natural openings do not arrive in the two minutes you have. Speak into a flat moment. It feels wrong and reads as normal.

“Just be yourself” is not advice, it is a refusal to give any. And drinking to get loose has a cost people underrate: it works on the current conversation and does nothing for the next sober one, so the skill never transfers.

Drop all four and you have already recovered most of the attention you were spending on them.

A Two-Week Practice Protocol That Risks Nothing

One exchange a day, with someone whose job it is to be there. A cashier, a barista, a receptionist, the person at the counter you were going to stand at anyway. The interaction was going to happen regardless, so the added risk is 15 seconds, and the rep is real because the other person is a stranger.

Keep the target low on purpose: one statement, one follow-up, done. Do not aim for a good conversation, aim for the sequence having been run. In everybody I have watched try this, the order came out the same and it was not the order they had braced for. What moved first was physical, the delay before speaking and the dread on the approach, and the fluency sat roughly where it had been. Ten or twelve reps in is where people tend to notice. That is an observation about the order of things, not a schedule, and there is no honest schedule to give.

Which brings up the instrument problem, and it is worth knowing about before the first rep rather than after the tenth. Boothby and colleagues named the liking gap in 2018 by measuring both sides of the same conversation: each partner rated the other person higher than they believed the other person had rated them. Two people, both wrong, wrong in the same direction, neither one able to tell.

The forecast has the same fault at the other end of the exchange. Epley and Schroeder ran commuter studies in 2014 in which riders assigned to strike up a conversation reported a better ride than riders assigned to keep to themselves, while a separate group asked to predict it in advance picked the opposite condition. Wrong going in, wrong coming out, and nothing in between corrects either error.

Count reps completed, then, not conversations that felt good. The second number is measured with a broken instrument, and the practice fails for most people at precisely the point where they start grading themselves on it.

FAQ

What do I say when my mind goes completely blank? Describe something you can both see, out loud. “This machine has never once worked on a Monday” or “these chairs are a punishment” require no memory search, because the content is in front of you. If nothing comes, saying “sorry, I lost my train of thought” is a normal sentence that people answer easily.

How do I make small talk without sounding fake? Say true things about small subjects rather than false things about big ones. The fakeness people detect comes from performed enthusiasm, not from a trivial topic, so a flat honest “long week” outperforms a bright “how ARE you.” Match the other person’s energy level within one step and you will not read as false.

How long does it take to get better at this? Nobody can give you an honest number, and anyone offering one is guessing. The order is the part worth watching: the dread on the approach eases before the fluency arrives, and holding something open past two exchanges is the last piece to land. Track which of those three has moved rather than counting weeks.

Is it worse to try and fail than not to try? Failing is cheaper than it feels, and the research pins the error at both ends: commuters pushed into talking with a stranger enjoyed the ride more than they had forecast, and people leaving a first conversation put the other person’s opinion of them below where that person actually put it. Assume your read on a dead exchange is wrong by the same margin, and stop replaying it.

What if I genuinely have nothing in common with them? You share the room, the wait, the weather, and whatever is on the table, which is enough material for the two minutes this needs to last. Common ground is what a conversation builds, not what it requires up front. Ask about the part of their situation you can see rather than hunting for an overlap that has not been established yet.

The version of this that helps is smaller than the version people imagine. You are not learning to be charming. You are learning to say an obvious true sentence about a shared room, take one word out of what comes back, and hand over something of your own the same size. That is the whole skill, and it is closer to a habit than a talent. Run it at a counter tomorrow morning and let the result be as uninteresting as it wants to be. Uninteresting is what success looks like at the start. The protocol above runs on reps, and the earliest ones can be even lower-risk than a cashier. Walking the four moves through Lona, an AI that keeps yesterday’s exchange in memory and that you can open any morning, lets you groove the describe-then-follow habit before you carry it to a counter. The counter is still the real rep. This only makes the approach less strange.

Sources

• Huang, K. et al., “It Doesn’t Hurt to Ask: Question-Asking Increases Liking,” Journal of Personality and Social Psychology, 2017
• Boothby, E. J. et al., “The Liking Gap in Conversations,” Psychological Science, 2018
• Epley, N. & Schroeder, J., “Mistakenly Seeking Solitude,” Journal of Experimental Psychology: General, 2014