Ice breakers that aren't cringe

Six in ten Britons who have been put through a workplace team building exercise call the experience "embarrassing or cringe-worthy", in YouGov's 2022 survey. Asked what they feel at the prospect of one, the most common answer, at 40%, is dread. And in a January 2025 Acas poll of 1,052 British employees, team building came out the single most disliked workplace social activity, at 31%, ahead of after-hours drinks.

Your dislike of ice breakers is the majority position, and it is measured.

The cringe has two causes, and neither is the question. The first is the label. The second is the format: one person at a time, out loud, with your name attached. Fix the label and fix the queue, and the opening becomes the most useful three minutes of the meeting.

Why calling it an ice breaker makes it worse

YouGov measured that dread at the prospect of a team building exercise, not during one. It is a reaction to an announcement. Say the words quick ice breaker and you have filed the next five minutes under a genre most of the room already dreads, before anyone has heard the question.

You have also announced two other things: that this is fun, and that management arranged it. That combination is taxed. Tews, Jolly and Stafford studied 491 employees across 141 stores. Fun activities went with lower staff turnover only where the fun read as less managed. Where it read as managed, they did nothing at all.

So never say the word. Describe the thing instead: everyone gets sixty seconds to draw, then we guess whose drawing is whose. A concrete description gives nobody anything to brace against.

Then hand the room a say. Patall, Cooper and Robinson's meta-analysis of 41 experiments found that choosing an activity, rather than being assigned it, raised intrinsic motivation, effort and task performance. Offer two prompts and let the room pick. It costs ten seconds and changes whose activity this is.

Why ice breakers feel so awkward

The standard opener is a queue: one person speaks, everyone else waits, and every answer arrives with a name on it.

Count what the queue costs. Fourteen people at twenty-five seconds each is nearly six minutes. Nobody spends the wait listening, because a turn is coming with their name on it, so they spend it drafting. A round-robin is not a group activity. It is fourteen solo performances in front of an audience that is mentally elsewhere.

Being the performer is worse than the room realises. Gilovich, Medvec and Savitsky sent people into a room wearing an embarrassing T-shirt, then asked how many observers had noticed. The wearers guessed exactly twice as many as actually had, a 23-point overestimate. The person on the spot is running a private worst case the room is not watching, because the room is rehearsing.

Attaching names finishes the job. Diehl and Stroebe told brainstormers that judges would rate their contributions, and the number of ideas produced dropped significantly, however the apprehension was created.

Sequential wastes the room. Spotlit taxes the speaker. Attributed shrinks the answers. None of the three is the question: what you would bring to a desert island is harmless on its own. Delivered through a queue, a spotlight and a name tag, it plays as an interrogation with catering.

Can you just skip it and start the meeting?

Skipping has a measured cost of its own. Allen, Lehmann-Willenbrock and Landowski surveyed 252 working adults and found small talk before a meeting predicted how effective people rated it, over and above good meeting procedures. The effect was strongest for the least extraverted people in the room. The unglamorous chat while people trickle in does the exact job staged openers claim to do for quiet people.

A warm-up also changes the work that follows it. Zheng and colleagues had pairs play an investment game that rewarded mutual trust. Among pairs given no warm-up, 35% never once fully cooperated. Among pairs who had simply chatted first, none.

The opening is load-bearing. What has to change is what it asks of people.

What works instead of ice breakers

Aim at something real first. Edmondson defines psychological safety as a shared belief that the team is safe for interpersonal risk taking. In her study of 51 work teams, safety reached performance through learning behaviour: asking, admitting, flagging problems early. That is what an opening can move. Not fun. Whether the next risky sentence in this meeting feels sayable.

The mechanism is Aron's. His team built four ingredients into the famous closeness procedure: escalating reciprocal disclosure, matching on attitudes, expected mutual liking, and naming closeness as the explicit goal. Tested element by element, only the task itself made a significant difference. Telling people to get close did nothing. The task is not weak medicine: pairs given the escalating questions beat matched small talk 4.06 to 3.25 on a closeness composite, a gap of .88 standard deviations.

That one result is the whole design brief. You cannot instruct a room to connect. You hand it a task whose by-product is connection, and you stay quiet about the by-product.

Put the evidence together and you get four constraints, enough to pick one activity instead of browsing two hundred. Everyone acts at the same moment. No name is attached to anything while it is being made. The task is real. And there is a right answer, so the room has a job to do together.

Why a drawing game beats a round of questions

Drawing and guessing satisfies all four constraints at once. Everyone draws the same prompt in the same sixty seconds. The drawings go up unsigned. The prompt is a real question about the person, not an instruction to bond. And guessing who drew which has a right answer, so the room has a job.

The drawing itself is not filler. In Wammes, Meade and Fernandes' experiments, drawing a word beat writing it for later recall at d = 0.67, and beat viewing a ready-made picture at 0.42. Four seconds per item was enough. The act of drawing is doing the work.

You cannot draw, and it does not matter. Fernandes, Wammes and Meade found the benefit unrelated to artistic ability, and it replicated in a lecture hall with 10 to 30 people drawing at once. A wobbly six-legged dog is better to guess about than a good one.

The prompt has to be specific, though. Boggs, Cohen and Marchand tested free-form doodling, and it scored significantly worse than structured shading and worse than note-taking. Free doodling is disengagement with a pen. Give the room something definite to draw, never a blank page.

The guessing is what turns parallel sketching into a group. A guess is a tiny theory about a colleague, tested in public, with none of the risk of a claim about yourself. The laughs at the reveal are load-bearing too. Lehmann-Willenbrock and Allen coded 54 videotaped team meetings and found humour patterns, a joke that draws laughter that draws another joke, predicted team performance immediately and two years later. Isolated jokes predicted nothing.

One rule is not negotiable: everyone plays. Galantucci and Garrod found onlookers who saw a pair's drawings but could not respond understood them far less well than the people in the game. Watching is not playing.

Guess the Doodler is this format, ready to run. Participants scan a QR code or open a short link and draw on their own phones; nothing installs and nobody makes an account. Drawings stay unattributed until the reveal, and a correct guess scores a flat 100 points, with no bonus for speed.

What to draw when the team already knows each other

For a team with years of shared history, names are not the gap. The map is. Atlassian surveyed 12,000 knowledge workers, and 1 in 2 said teams at their company unknowingly work on the same things. Half.

Groups are reliably bad at surfacing what they collectively know, even when paid to. Lightle, Kagel and Arkes' groups uncovered only 35% of hidden profiles when 75% was achievable on recall alone. The information was in the room. The room did not produce it.

So make the prompt the work. Give the room "Draw your area of responsibility". Sixty seconds, phones out, no signatures. The guessing round that follows is a live test of the team's map of itself, and the wrong guesses are the finding. When three people file the billing pipeline under the wrong engineer, the team has just learned why work gets duplicated, and it learned it laughing. Say the surprising misattribution out loud while the drawing is still up; that sentence is worth more than the scoreboard.

The sillier variant, "Draw your job title", runs well in rooms that already laugh easily. Responsibility asks the question the org chart does not answer.

What to draw when nobody knows each other

Cable, Gino and Staats rebuilt the first day at Wipro around the new hires themselves rather than the company. Across 605 people, that made them more than 32% less likely to quit within six months than standard onboarding. A first-day session about the organisation's identity also lost to it. The content that binds a newcomer is the person, not the company.

So the prompt for a room of strangers is "Draw your hidden talent". It asks for exactly what the Wipro result says to ask for: a true fact about the person. The drawing does the disclosing, so nobody has to narrate themselves to a silent room. For a genuinely cold room, a day-one hiring cohort or two departments meeting after a merger, the gentler prompt is "Draw your past weekend". Offer both out loud and let the room pick by show of hands.

Kardas, Kumar and Epley found people overestimate how awkward getting acquainted will be, and overestimate it most for the deepest prompts. The dread is real. The event is smaller than the dread.

Should you play Two Truths and a Lie?

Skip it, and the best argument against it comes from someone who recommends it.

Gelman teaches the game and published a paper on it. In his own applied-regression class of 49, the students picked the lie correctly 23 times, which he describes as not much better than the one-in-three you get by guessing at random. In the same paper he writes that there is no direct empirical evidence of the activity's effectiveness on student learning. That is a supporter's own assessment, in a peer-reviewed paper.

The format also selects for bragging. It invites impressive personal facts, and Scopelliti, Loewenstein and Vosgerau found self-promoters overestimate how proud and happy listeners will feel and underestimate how annoyed they will be.

The last objection is a judgement rather than a finding, and it is worth marking as one. The game asks a new colleague to lie convincingly to a room that is forming its first impression of them. The nearest research studies unexpected lying between strangers with money at stake. In that setting Schweitzer, Hershey and Bradlow found trust damaged by deception never fully recovered, not after a promise, an apology, or a consistent run of trustworthy actions. A party game where everyone knows one statement is false is not that setting. Treat it as a reason to be careful, not as a result.

Run as a silent quiz, with the statements on screen and everyone voting at once from phones, the game genuinely fixes the queue. That is a real improvement, and it is the only one. The content is still a lie, told with your name on it, to people forming their first impression of you.

Dropping it costs nothing, because the closeness the game borrows comes from truthful reciprocal disclosure, not from the deception. That was Aron's whole result. A drawing keeps everything worth keeping. There is still a secret, still a vote, still a scoreboard, still a reveal. The one thing missing is the lie.

And if psychological safety was the deeper goal, the strongest evidence points somewhere else. The one randomised trial at scale, across more than 1,000 teams, raised safety by 16% of a standard deviation through managers holding frequent one-to-one conversations about individual needs. No game was involved.

How to run it, and what to do when it lands flat

Sixty seconds of drawing is enough, and the rest of the opening is a handful of decisions most facilitators make by accident.

Length. The default drawing time in Guess the Doodler is 60 seconds, and the presets run 30, 60, 90 and 120. Short is correct. Chopik and Oh report closeness effects in earlier work from openings as brief as 6, 12 and 30 minutes. A full drawing round with its reveal fits inside five minutes. The catch is the reveal: every submitted drawing joins the queue, so the round count equals the head count. In a forty-person town hall, reveal a handful and then end the activity, or give it a slot of its own.

Order. Before the agenda, not as a mid-meeting break. The Allen small-talk effect was measured before the meeting started, and Zheng's warm-up preceded the game it changed. Openers work by changing what follows them.

Guessing. Keep it on phones. Fay, Garrod and Carletta found five-person groups talk like a dialogue while ten-person groups follow whoever talks most, and Guess the Doodler runs one undivided room. Simultaneous silent guessing is the one format the loudest voice cannot dominate.

Everyone plays. Exempt a few people and you have rebuilt the audience the format exists to remove, and Galantucci's onlookers were the ones who understood the drawings least.

Cameras, hybrid and dial-in. A camera is not participation. Shockley and colleagues ran a four-week field experiment, 1,408 daily observations from 103 employees, and camera-on raised daily fatigue, which related negatively to speaking up. The effects were larger for women and newer employees. Drawing on a phone is participation that never asks for a face. It is also the answer for the conference room with six remote joiners: everyone plays from their own phone, so the remote people get the identical game, not a degraded feed. The audio-only case, someone dialling in from a car, is where the game dies. Do not run a drawing game at drivers; give that meeting two unstructured minutes before the agenda instead.

Prompts. Concrete and outward-facing: a weekend, a talent, a responsibility. Never anything that asks for a confession.

When not to run it. With three people or fewer, authorship is obvious and the game collapses; just talk. When the meeting is bad news, open with the bad news: an activity in front of redundancies reads as management deciding how the room should feel, and everybody can tell. And in a daily standup with the same five faces, retire it before it stales. An opener run by reflex is the label growing back.

Sometimes the round lands flat anyway. Remember the T-shirt study before you panic: ninety flat seconds at the front of a room feels like catastrophe and reads, from the seats, as Tuesday. Never apologise for the activity. An apology confirms there was something to apologise for.

If the room goes quiet at a reveal, make one specific wrong guess yourself: this one is Maya's, because of the spreadsheet with teeth. A specific guess is cheap to contradict, contradicting is participation, and a presenter being wrong is the cheapest proof that being wrong here costs nothing.

If one person is answering everything, ask for a second guess before you reveal. A wrong guess from somebody else is what turns it back into a game.

If someone is sarcastic, agree lightly and keep going. The format is silly, the drawings are the point, and the round is shorter than the complaint.

If submissions stall, end the drawing phase early. A round that finishes while people still want one more beats a long one you had to rescue.

And when a drawing gets zero correct guesses, name one thing it got right, ask the author one concrete question about it, and move on. The next drawing is thirty seconds away, which is a recovery mechanism no speech has ever had.