Creativity training works. Just not as well as anyone told you.

How to run it

Give a divergent prompt — a real design problem, or the classic Alternate Uses Task ("list uses for a brick"). Set a ten-minute minimum and enforce it. No stopping at "I'm done." No evaluation, no editing, quantity only.

Afterwards, have each person circle their three most original ideas and note where in the list they fall. Show of hands: how many circled something from the back half?

Facilitator script

1. 2 min · Set up
   For the next ten minutes there is exactly one rule: quantity. No editing, no judging, no crossing out. If an idea is stupid, write it down anyway — stupid ideas are load-bearing.
2. 10 min · Silent generation
   Ten minutes on the clock. When you feel like you have run out, that is the halfway point, not the end. Keep going.
3. 3 min · Self-selection
   Now go back through your list and circle the three you think are most original. Not most useful — most original.
4. 8 min · The reveal
   Hands up if any of your three came from the first half of your list. Now the second half. Look around the room. That gap is one of the oldest findings in creativity research, and you just reproduced it in ten minutes.
5. 2 min · Land it
   The implication is uncomfortable. Every brainstorm you have ever cut short at the natural energy dip ended precisely before the good ideas arrived.

The evidence

The serial order effect is one of the oldest and most robust findings in creativity research, dating to Christensen, Guilford & Wilson (1957). Beaty & Silvia (2012) time-stamped every response from 133 adults across a ten-minute unusual-uses task and had three raters score each for creativity. Originality rises across time while fluency declines. It has replicated in adults, older children, and five-to-six-year-olds.

Early responses come from direct memory retrieval — known uses, seen uses. Only once that well runs dry do people engage the effortful executive processes that produce genuinely remote associations. The EEG picture matches: alpha power shows a U-shaped curve, dipping mid-task before rising again ahead of the more original responses.

Evidence profile

Cognitive-skills ideation training is the moderator the 169-study meta-analysis supports; outcome is divergent-thinking scoring, not real output.

Beaty, R. E., & Silvia, P. J. (2012). Why do ideas get more creative across time? Psychology of Aesthetics, Creativity, and the Arts, 6(4), 309–319. doi:10.1037/bul0000432 Find this paper ↗

How to run it

Split the group. Half receive the open brief ("design a better commute"). Half receive the same brief plus an arbitrary constraint — "and it must involve a bucket", "using only things already in this building", "in under fifty words".

Collect outputs, strip the condition labels, and have a third party rate them for creativity. Reveal the split afterwards.

Facilitator script

1. 3 min · Distribute
   You each have a brief. Do not compare them yet. Some of you have an extra line — that is deliberate, and I will explain at the end.
2. 12 min · Generate
   Twelve minutes. Work individually. Whatever is on your sheet is the whole task.
3. 8 min · Blind rating
   Pass your sheet to the far side of the room. Score each one you receive from one to five on originality only. Do not score usefulness.
4. 8 min · Reveal and discuss
   Now compare briefs with the person next to you. Which group scored higher? Almost every time, it is the group who had less freedom.
5. 4 min · Transfer
   So the practical question is not how do we remove obstacles. It is which constraint do we choose, and do we choose it deliberately or let the org chart choose it for us.

The evidence

Haught-Tromp (2017), Psychology of Aesthetics, Creativity, and the Arts: participants wrote two-line greeting-card rhymes. On half the trials they were additionally required to include a specific, arbitrary noun. Constrained rhymes were rated more creative, in both studies. An order interaction also indicated a carryover effect — practice with constraints improved performance on later unconstrained tasks.

Evidence profile

Controlled experiments plus a meta-analytic review of the inverted-U; outcomes are rated idea originality.

Haught-Tromp, C. (2017). The Green Eggs and Ham hypothesis: How constraints facilitate creativity. Psychology of Aesthetics, Creativity, and the Arts, 11(1), 10–17. Find this paper ↗

How to run it

State the problem as an abstract function, stripping every piece of domain vocabulary. "How do we onboard new employees faster" becomes "how does a system rapidly integrate a foreign element without rejecting it."

Then ask who else solves that — immunology, coral reefs, jazz ensembles, aviation, prison systems, beekeeping. Harvest three mechanisms from each distant domain and translate them back.

Facilitator script

1. 8 min · Strip the language
   Write your problem on the board. Now remove every noun specific to your industry. If a stranger could not tell what business you are in from the sentence, you have done it right.
2. 5 min · Draw domains
   Each table draws three domain cards. You do not get to swap. The further from your world, the better this works.
3. 15 min · Mine mechanisms
   How does your domain solve the abstract problem? I want mechanisms, not vibes. Not "bees are collaborative" — how does a hive actually allocate foragers to a food source?
4. 12 min · Translate back
   Now map it. What is the equivalent of the waggle dance in your onboarding process? Be literal. The literalness is where the idea is.
5. 5 min · Debrief
   Notice which analogies produced something and which produced decoration. The difference is always whether you mapped function or surface.

The evidence

Jeppesen & Lakhani (2010), Organization Science, analysed 166 R&D challenges broadcast to over 12,000 scientists on InnoCentive. The probability of submitting the winning solution rose with increasing distance between the solver's field of expertise and the problem's field. Outsiders won not despite their distance but because of it — they imported perspectives and heuristics unavailable inside the field.

Evidence profile

Multiply replicated in analogical-transfer experiments; the field-outcome evidence sits in activity 09 rather than here.

Jeppesen, L. B., & Lakhani, K. R. (2010). Marginality and problem-solving effectiveness in broadcast search. Organization Science, 21(5), 1016–1033. Find this paper ↗

How to run it

Front-load the work: 20–30 minutes of genuine, effortful attempt at the problem. Then break for 15–20 minutes with a low-cognitive-demand task — a walk, tidying, undemanding admin. No email, no meetings, no problem-adjacent thinking. Return and re-attempt.

Facilitator script

1. 25 min · Load the problem
   Twenty-five minutes of real attempt. Not planning to attempt — attempting. You need to hit a wall for the next part to work at all.
2. 18 min · Low-demand break
   Now put it down completely. Go for a walk, tidy something, do the dishes. Not your phone. Not your inbox. The task needs to occupy your hands and almost none of your mind.
3. 12 min · Re-attempt cold
   Straight back in, no warm-up, no recap. Write for two minutes before you speak to anyone.

The evidence

Sio & Ormerod (2009), Psychological Bulletin, meta-analysed 117 studies. Overall incubation effect: d = 0.29. Three moderators matter for how you run it. Divergent thinking tasks benefit most, more than linguistic or visual insight problems. A low-demand break task beats a high-demand task and beats resting with no task at all. And longer preparation before the break produces a larger effect.

Evidence profile

Meta-analysis of incubation studies with moderator analysis; outcomes are lab divergent and insight tasks.

Sio, U. N., & Ormerod, T. C. (2009). Does incubation enhance problem solving? A meta-analytic review. Psychological Bulletin, 135(1), 94–120. doi:10.1037/a0014212 Find this paper ↗

How to run it

Move divergent thinking sessions onto their feet. Either a corridor loop or an outdoor walk. Pairs work well: one walks and talks, the other captures. If you cannot walk during the session, walk immediately before it — the effect persists.

Sit down again before any decision-making.

Facilitator script

1. 2 min · Pair and brief
   Pair up. One of you talks, one of you writes — and the writer does not edit, does not improve, does not filter. You are a recording device with legs.
2. 8 min · Walk, person A
   Eight minutes. Keep moving the whole time. If the conversation stalls, keep walking through the silence rather than stopping.
3. 8 min · Walk, person B
   Swap. Same rules.
4. 2 min · Sit down
   Now sit. Deliberately. We are switching from generating to choosing, and the research says those want different physical states.

The evidence

Oppezzo & Schwartz (2014), Journal of Experimental Psychology: LMC, four experiments, 176 participants. Walking increased divergent thinking on Guilford's Alternate Uses for 81% of participants in Experiment 1; across the four experiments the proportion of walkers who improved was 81%, 88%, 95% and 100%. The effect held indoors on a treadmill facing a blank wall as well as outdoors, so it is not environmental stimulation. There was a residual effect: people who sat down after walking still outperformed people who sat throughout.

Critically, walking did not help convergent thinking — Compound Remote Associates performance was mildly worse.

Evidence profile

Four within-subject experiments with control conditions, independently extended; outcome is AUT and analogy generation.

Oppezzo, M., & Schwartz, D. L. (2014). Give your ideas some legs. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40(4), 1142–1152. doi:10.1037/a0036577 Find this paper ↗

How to run it

Best taught as a demonstrated technique participants try at home rather than a workshop activity.

Load the problem deliberately. Sit semi-reclined holding a light object over a hard surface. Doze. When the object falls, immediately write down whatever was in your mind — before you evaluate any of it.

Facilitator script

1. 5 min · Load
   Spend five minutes deliberately turning the problem over. You are not trying to solve it. You are trying to make it the most recently active thing in your head.
2. 15 min · Drift
   Semi-recline, object in hand over a hard floor. Let yourself go. The drop will wake you.
3. 5 min · Capture immediately
   Write before you judge. Hypnagogic material reads as nonsense on the way in and turns out to be a metaphor on the way out. Judge it tomorrow.

The evidence

Lacaux et al. (2021), Science Advances, N = 103. Participants were given maths problems containing a hidden shortcut rule. During a 20-minute rest with EEG monitoring, those who spent as little as 15 seconds in N1 — the hypnagogic sleep-onset stage — discovered the rule at a rate of 83% versus 30% for those who stayed awake. The benefit disappeared if participants slipped past N1 into deeper N2 sleep.

Evidence profile

Well-controlled polysomnography study with a clean comparison, but a single sample and a rule-discovery outcome rather than open ideation.

Lacaux, C., Andrillon, T., et al. (2021). Sleep onset is a creative sweet spot. Science Advances, 7(50), eabj5866. doi:10.1126/sciadv.abj5866 Find this paper ↗

How to run it

1. Silent generation — ten minutes, individually, in writing
2. Round-robin capture — each person reads one idea at a time, no discussion, until all are on the wall
3. Clarify — questions of understanding only, no evaluation
4. Build — now go verbal, using the pooled set as raw material
5. Converge — see activity 12

Facilitator script

1. 2 min · Set the rule
   We are not going to brainstorm out loud, and I will tell you why afterwards. For now: silence, and one idea per card.
2. 10 min · Silent generation
   Ten minutes. One idea per card. Do not talk, do not look at anyone else's cards, do not stop early.
3. 12 min · Round-robin
   Going round the circle, one card each, read it and put it up. No commentary, no "oh I had that too", no reactions. We are building the pool, not judging it.
4. 6 min · Clarify only
   Questions of understanding only. You may ask what someone meant. You may not ask whether it would work.
5. 10 min · Now talk
   Now the group conversation is worth having, because everybody's ideas are already in the room instead of trapped behind whoever spoke first.

The evidence

Diehl & Stroebe (1987) ran four experiments isolating three candidate mechanisms for the productivity loss in brainstorming groups — free riding, evaluation apprehension, and production blocking — and found that production blocking accounted for most of it. Mullen, Johnson & Salas (1991) meta-analysed the literature and confirmed that nominal groups reliably outproduce interacting brainstorming groups, with the gap widening as group size grows.

Production blocking is mechanical. Only one person can speak at a time, so while you wait your turn you either forget your idea, suppress it as no-longer-relevant, or lose your thread listening to someone else. In a six-person group each member spends the large majority of the session listening rather than generating.

Evidence profile

Meta-analytic evidence on production blocking and nominal-versus-real groups, replicated for four decades.

Diehl, M., & Stroebe, W. (1987). Productivity loss in brainstorming groups. Journal of Personality and Social Psychology, 53(3), 497–509. doi:10.1037/0022-3514.53.3.497 Find this paper ↗

How to run it

Ban solutions for the first thirty minutes. Instead, generate problem statements. Take the presenting problem and rewrite it at five levels of abstraction, then generate at least ten alternative framings. Only then vote on which framing to ideate against.

Facilitator script

1. 3 min · Impose the ban
   For the next half hour, nobody proposes a solution. If a solution arrives, write it on a card, put it face down, and come back to it later. It will still be there.
2. 10 min · Climb the ladder
   Take your problem and ask "why does that matter?" five times, writing each answer as a new problem statement. You are climbing towards the abstract.
3. 10 min · Ten framings
   Now ten alternative framings, anywhere on that ladder. Some should feel too broad and some too narrow. That is the range we want to see.
4. 7 min · Choose deliberately
   Vote. And notice what you are doing — you are choosing the shape of every idea you will have for the rest of the day.

The evidence

This is the practical expression of the strongest moderator in the Sio & Lortie-Forgues meta-analysis: cognitive-content training outperformed non-cognitive training, and this survived the full multi-moderator model (p = .024). Problem construction is the most cognitive of the creativity skills — it is the part where an actual thinking procedure is taught. It maps onto Mumford's problem-construction research programme, which consistently finds that how a problem is defined constrains the entire downstream solution space.

Evidence profile

Controlled problem-construction experiments, independently replicated; outcomes are rated solution quality and originality.

Sio, U. N., & Lortie-Forgues, H. (2024). The impact of creativity training on creative performance. Psychological Bulletin, 150(5), 554–585. Find this paper ↗

How to run it

When assembling an ideation team, deliberately include at least one person with no domain expertise in the problem, and brief them explicitly that naive questions are their job, not a liability. Pair them with a domain expert acting as translator rather than gatekeeper.

Give the outsider a formal right of interruption. Without it, social pressure silences them inside ten minutes.

Facilitator script

1. 5 min · Brief the outsider privately
   Your job today is to ask the questions everyone else is too embarrassed to ask. If you understand everything being said, you are not doing your job.
2. 2 min · Announce the role
   Sam is here as our outsider. Sam has a standing right to stop us at any point. When Sam does, we answer properly — no shorthand, no acronyms.
3. 8 min · Harvest at the end
   Sam, which of our assumptions sounded strangest to you? Not wrong — strange. Those are the ones worth checking.

The evidence

Jeppesen & Lakhani (2010) as above. Also relevant: Maddux & Galinsky (2009), JPSP, found across five studies that time spent living abroad predicted creative performance, that the effect was causal, and that it was mediated by degree of cultural adaptation rather than mere exposure. Maddux, Adam & Galinsky (2010) sharpened this: the active ingredient is functional multicultural learning — understanding why people in another culture do what they do, not just observing that they do it.

Evidence profile

Large field dataset with real winning solutions, but observational — solvers were not randomly assigned a distance from the domain.

Maddux, W. W., & Galinsky, A. D. (2009). Cultural borders and mental barriers. Journal of Personality and Social Psychology, 96(5), 1047–1061. doi:10.1287/orsc.1090.0491 Find this paper ↗

How to run it

Standard, well-tested levers: leaders model fallibility by naming their own errors first; frame the work explicitly as a learning problem rather than an execution problem; respond to bad news with curiosity rather than blame; separate idea generation from idea ownership so that critique cannot be read as personal attack.

Facilitator script

1. 10 min · Leader goes first
   I am going to start by describing a call I got wrong this quarter, and what it cost. Nobody else speaks until I am finished.
2. 15 min · Reframe the work
   Is this a problem where we already know the answer and need to execute, or one where we do not and need to learn? Be honest. Most teams claim the second and behave like the first.
3. 20 min · Detach ideas from owners
   From here on, ideas go on the wall unattributed. You may attack any idea on that wall as hard as you like. You may not attack a person.
4. 15 min · Test it
   Someone name something that is going worse than we have been saying out loud. I will respond with questions, not consequences. Watch what I do — that is the actual policy.

The evidence

Frazier et al. (2017), Personnel Psychology, conducted a meta-analytic review confirming that psychological safety significantly predicts creativity, voice behaviour, information sharing, learning behaviour and task performance, at both individual and team levels. Edmondson (1999) established the construct; Newman, Donohue & Eva (2017) systematically reviewed the literature and mapped its antecedents.

A necessary nuance: recent work on nonlinear relationships finds the benefits concentrated in non-routine, exploratory work. For highly standardised tasks the returns flatten. Psychological safety is a creativity intervention, not a universal performance intervention.

Evidence profile

Field studies and meta-analysis with real team outcomes; the design is predominantly correlational rather than experimental.

Frazier, M. L., et al. (2017). Psychological safety: A meta-analytic review and extension. Personnel Psychology, 70(1), 113–165. doi:10.2307/2666999 Find this paper ↗

How to run it

A twenty-minute teaching module, ideally delivered to decision-makers rather than ideators. Present the finding, then audit your own organisation's last five rejected proposals against it.

Then introduce structural counters: blind evaluation, mandated advocacy for the most novel option, and separating the uncertainty-reduction conversation from the novelty-assessment conversation.

Facilitator script

1. 5 min · Present the finding
   People reject novel, high-quality ideas while sincerely reporting that they want creative ideas. Not hypocrisy — an implicit bias that switches on under uncertainty.
2. 10 min · Audit your own
   Pull up the last five proposals this group turned down. For each one, ask: was the stated reason about the idea, or about our uncertainty? Be uncomfortable about this.
3. 5 min · Install a counter
   From today, every shortlist carries one high-novelty option with a named advocate whose job is to argue for it. Not to believe in it — to argue for it.

The evidence

Mueller, Melwani & Goncalo (2012), Psychological Science. Two experiments manipulating uncertainty, including an uncertainty-reduction prime. Results showed a negative implicit bias against creativity relative to practicality whenever participants experienced uncertainty — and, more damagingly, this bias impaired participants' ability to recognise a creative idea when they saw one.

Evidence profile

Randomised experiments with implicit measures, independently replicated; outcome is evaluation of creative ideas, not production.

Mueller, J. S., Melwani, S., & Goncalo, J. A. (2012). The bias against creativity. Psychological Science, 23(1), 13–17. doi:10.1177/0956797611421018 Find this paper ↗

How to run it

1. Cluster the raw pool by underlying mechanism, not surface theme
2. Score on two independent axes — novelty and feasibility — separately, and by different people
3. Force-select from the novelty quadrant. At least one advancing concept must come from high-novelty / low-feasibility, with a named advocate
4. Prototype to learn, not to prove. The first prototype converts an uncertainty into a cheap question
5. Record the rejected set. It is the most valuable artefact of the session and is almost always binned

Facilitator script

1. 12 min · Cluster by mechanism
   Group these by how they work, not by what they are about. Two ideas in different markets that use the same mechanism belong together.
2. 10 min · Score novelty only
   This half of the room scores novelty. Only novelty. I do not want to hear the word "realistic".
3. 10 min · Score feasibility only
   This half scores feasibility. Only feasibility. I do not want to hear the word "exciting".
4. 8 min · Plot and force-select
   Plot them. Now — one advancing concept has to come from the top-left quadrant. Who is advocating for it?
5. 5 min · Bank the rejects
   Photograph the whole wall including everything we killed. In six months the constraint that killed half of these will have changed.

The evidence

Divergent thinking is only half of creativity — the definition used across this literature is ideas that are novel and useful. Yet most creativity activities, and most creativity research, address only novelty: in the Sio & Lortie-Forgues sample, 137 of 169 studies used divergent thinking as the outcome and only 29 measured an actual creative product.

Research on idea selection consistently finds that groups choose ideas that are feasible and familiar over ideas that are original — even when explicitly instructed to select the most creative option, and even when they generated the original ideas themselves.

Evidence profile

Controlled selection experiments; the specific protocol here is assembled from several studies rather than tested as a whole.

Rietzschel, E. F., Nijstad, B. A., & Stroebe, W. (2010). The selection of creative ideas after individual idea generation. British Journal of Psychology, 101(1), 47–68. Find this paper ↗

How to run it

Teach the room to name which mode they are in, out loud, before speaking. Give each person two pens — one colour for generative contributions, one for evaluative. Everything written must use the matching colour.

Crude, slightly silly, and extremely effective at making the switching visible.

Facilitator script

1. 3 min · Explain the architecture
   Generating and evaluating run on brain networks that are anticorrelated at rest. When I ask you to "be creative but realistic", I am asking you to run two opposed systems at once. That is not rigour. That is a design fault in the instruction.
2. 2 min · Assign the pens
   Blue is generative. Red is evaluative. You may switch as often as you like — but you have to pick up the other pen to do it.
3. 5 min · Run and observe
   Look at the balance of colour on your sheet. Most teams discover they are running eighty percent red and calling it ideation.

The evidence

Beaty et al. (2018), PNAS: connectome-based predictive modelling on 163 participants identified a network spanning default, salience and executive systems — networks that typically work in opposition. Connectivity strength within it predicted idea originality across four independent datasets.

A 2025 multi-centre study in Communications Biology (10 samples, five countries, N = 2,433) found that the number of dynamic switches between default and executive networks predicted creativity but not general intelligence.

Evidence profile

Converging evidence from group ideation experiments and the network-switching neuroimaging literature.

Beaty, R. E., Kenett, Y. N., et al. (2018). Robust prediction of individual creative ability from brain functional connectivity. PNAS, 115(5), 1087–1092. Find this paper ↗

How to run it

For the generation phase only: dim the lights, close laptops, stop talking, and offer explicit permission to close eyes. No music, no ambient video, no facilitator narration over the silence.

Then bring the light and energy back for the sharing phase. The contrast is the point.

Facilitator script

1. 2 min · Change the room
   Laptops closed, lights down, phones face down and out of reach. If you want to close your eyes for this, close them.
2. 8 min · Generate inward
   I am going to stop talking now. I will not fill the silence, and neither should you. Eight minutes.
3. 5 min · Bring it back up
   Lights up. Now we go loud. Notice how different that felt from a normal workshop — and that the difference was mostly the absence of me talking.

The evidence

EEG alpha power (8–13 Hz) rises reliably during creative idea generation and rises more in more creative individuals; Fink & Benedek's review describes it as among the most consistent findings in creativity neuroscience. Benedek et al. showed alpha increases specifically when bottom-up processing is prevented, supporting the interpretation of alpha as internally-directed attention — the brain damping external input to work with internal material.

That was correlational until Ritter, Abbing & van Schie (2018) tested it directly. Participants completed creativity and working-memory tasks twice, once with eyes open and once with eyes closed. Eye-closure improved performance on both the divergent task (Adapted Alternative Uses) and the convergent task (Remote Associates) — with no effect on working memory. The absence of a working-memory effect is the important part: it indicates the benefit is specific to creativity rather than a general cognitive boost.

Evidence profile

EEG correlational evidence plus a controlled eye-closure experiment; see activity 38, which points the other way.

Ritter, S. M., Abbing, J., & van Schie, H. T. (2018). Eye-closure enhances creative performance on divergent and convergent creativity tasks. Frontiers in Psychology, 9, 1315. See also Fink, A., & Benedek, M. (2014), Neuroscience & Biobehavioral Reviews, 44, 111–123. Find this paper ↗

How to run it

A planning rule rather than an activity. Poll the group's chronotypes in advance, then schedule the divergent session against the grain — late afternoon for a morning-heavy team, early morning for an evening-heavy one.

Keep the sharp hours for analytic work and decisions, which time of day does not reliably help anyway.

Facilitator script

1. 2 min · Poll in advance
   Before the session: are you sharpest first thing, or later in the day? One word answers, in the invite.
2. 3 min · Explain on the day
   You may have noticed I have scheduled our idea generation at four in the afternoon, which feels like a strange choice. It is deliberate, and the reason is that slightly foggy is better for insight than razor sharp.

The evidence

Wieth & Zacks (2011), Thinking & Reasoning: participants solved insight and analytic problems at their optimal or non-optimal time of day. Insight problem-solving was consistently better at the non-optimal time. Analytic problem-solving showed no consistent time-of-day effect. The proposed mechanism is reduced inhibitory control — the same loosening that lets irrelevant material into working memory also lets remote associations in.

Evidence profile

Single controlled experiment with a clean crossover design; replication record is thin and mixed.

Wieth, M. B., & Zacks, R. T. (2011). Time of day effects on problem solving: When the non-optimal is optimal. Thinking & Reasoning, 17(4), 387–401. doi:10.1080/13546783.2011.625663 Find this paper ↗

How to run it

A 15–20 minute guided body-scan or Yoga Nidra track between the generation and convergence phases, lying down, eyes closed.

If you run it, be straight with the room about the evidence — which is exactly the teaching moment this activity is best used for.

Facilitator script

1. 2 min · Frame it honestly
   We are going to do twenty minutes of guided deep rest. I want to be upfront: the famous dopamine finding behind this comes from eight people in a scanner in 2002, and creativity was never measured. We are doing it because a low-demand break is well evidenced, and this is a pleasant way to take one.
2. 18 min · Run the track
   Lie down, close your eyes, and follow the audio. If you fall asleep that is fine — although the research on sleep-onset suggests you want to stay just this side of it.

The evidence

Kjaer et al. (2002), Cognitive Brain Research. Eight experienced yoga teachers underwent ¹¹C-raclopride PET scans during Yoga Nidra. Binding in the ventral striatum fell 7.9%, which the authors calculate corresponds to a 65% increase in endogenous dopamine release — correlated with a concomitant rise in EEG theta activity. It was the first in vivo demonstration linking endogenous neurotransmitter release to conscious experience, and it is careful work.

It is also n = 8, in expert practitioners, measuring dopamine rather than any cognitive outcome. The step from "dopamine rose" to "you will be more creative" is an inference layered on by others, not a result.

Evidence profile

Small expert-sample PET study. Creativity was never measured — the link to creative output is an inference, not a finding.

Kjaer, T. W., Bertelsen, C., Piccini, P., Brooks, D., Alving, J., & Lou, H. C. (2002). Increased dopamine tone during meditation-induced change of consciousness. Cognitive Brain Research, 13(2), 255–259. Find this paper ↗

How to run it

Where programme design allows: a multi-day, device-free offsite in a natural setting before a major creative push.

Where it does not: protected, unstructured outdoor time with phones surrendered, framed as work rather than a break. The framing matters — people will not use it properly if it is billed as downtime.

Facilitator script

1. 5 min · Collect the devices
   Phones in the box. Not on silent — in the box. This is the intervention, not the inconvenience around the intervention.
2. 90 min · Unstructured outdoor time
   No agenda, no deliverable, no pairing. Walk, sit, whatever you like. If you find yourself solving the problem, let it happen — but do not go looking.
3. 20 min · Cold capture on return
   Before you talk to anyone, write for five minutes. What surfaced?

The evidence

Atchley, Strayer & Atchley (2012), PLOS ONE. 56 Outward Bound participants on four-to-six-day wilderness expeditions with no electronic devices. 24 took the Remote Associates Test the morning before departure (mean 4.14 of 10); 32 took it on the fourth morning of the trip (mean 6.08) — roughly a 50% improvement. The authors invoke Attention Restoration Theory: modern environments are full of sudden events that hijack attention, while natural settings offer "soft fascination" allowing the executive attention system to replenish.

Evidence profile

Small field study without an active control; outcome is the Remote Associates Test.

Atchley, R. A., Strayer, D. L., & Atchley, P. (2012). Creativity in the wild. PLOS ONE, 7(12), e51474. Find this paper ↗

How to run it

Audit the innovation calendar. Separate work into exploration blocks — unbounded, protected, no deliverable — and delivery blocks. Never schedule ideation into the gaps between deadlines.

Run it as a live exercise with the actual calendar on screen. It is considerably more confronting that way.

Facilitator script

1. 10 min · Put the calendar up
   This is last month, real. Point to the block where creative work was supposed to happen. Now count the interruptions inside it.
2. 12 min · Separate the two kinds
   Colour every block: is this exploration or delivery? Anything that is both is actually delivery.
3. 8 min · Defend one block
   Pick one exploration block a week and make it genuinely uninterruptible for a month. One. Then we look at what changed.

The evidence

Amabile's componential model and her diary-study research on time pressure found creative thinking was generally lower on high-time-pressure days. The notable exception: pressure did not damage creativity when people felt they were on a meaningful mission, were protected from interruption and distraction, and could focus on a single activity.

Evidence profile

Diary study of real projects with real creative output, replicated in related field work; observational, so causal direction is not established.

Amabile, T. M., Hadley, C. N., & Kramer, S. J. (2002). Creativity under the gun. Harvard Business Review, 80(8). Find this paper ↗

How to run it

You will meet this one in the wild: thirty seconds of looking left and right boosts creativity by connecting the two sides of your brain. The underlying paper exists and is properly published. Almost everything added to it in transmission does not hold.

Rather than running the technique, run the trace. Hand out the 2009 abstract and the 2015 replication abstract. Ask the room to identify, in fifteen minutes, exactly which claim was made by whom — and where the sentence "connects the two hemispheres" actually entered the story.

Facilitator script

1. 3 min · State the popular version
   Here is a claim you will have seen on LinkedIn. Thirty seconds of horizontal eye movements increases your ideation, because it connects the two sides of your brain. Hands up if that sounds plausible.
2. 7 min · Read the original abstract
   This is the actual 2009 paper. Read it and tell me three things: who benefited, on which measures, and whether the hemisphere explanation is presented as a finding or as a hypothesis.
3. 8 min · Read the replication
   Now the 2015 paper. Note the design — proponents and sceptics agreed the method together, preregistered it, and committed to publish either way. What did they find, and what did it do to the mechanism?
4. 7 min · Generalise the lesson
   The paper was fine. The journal was fine. The failure happened entirely in transmission, between the abstract and the slide. That is the pattern you should now expect by default — and the reason every card on this site names its primary source.

The evidence

Shobe, Ross & Fleck (2009), Brain and Cognition, tested 62 participants on the Alternate Uses Task after a 30-second bilateral eye movement task versus a central-fixation control. Bilateral eye movements raised originality and categorical distinctiveness in strong-handers only; mixed-handers showed no effect from the manipulation. Fluency, appropriateness and detail were unaffected. So the honest summary is that two of five sub-scores improved, in roughly half the sample.

The interhemispheric interaction account was always the proposed mechanism, not a result. It has since taken heavy damage. Matzke et al. (2015) ran a preregistered adversarial collaboration — proponents and sceptics agreed the design in advance and committed to publish regardless — and failed to replicate the eye-movement benefit on memory, with strong Bayesian evidence for the null. Roberts, Fernandes & MacLeod (2020) found the same pattern in PLOS ONE: a weak replication in Experiment 1, then substantial Bayesian evidence for a null in a larger, better-powered Experiment 2. Samara et al. (2011) had already found no consistent change in interhemispheric EEG coherence following bilateral saccades across six frequency bands.

None of that formally overturns the creativity result, which was a different task. It does dismantle the explanation everyone repeats.

Evidence profile

Competent original experiment whose proposed mechanism failed a preregistered adversarial replication.

Shobe, E. R., Ross, N. M., & Fleck, J. I. (2009). Influence of handedness and bilateral eye movements on creativity. Brain and Cognition, 71(3), 204–214. See also Matzke et al. (2015), JEP: General, 144(1), e1–e15. Find this paper ↗

How to run it

Write the brief twice. Version A is how you would normally phrase it, full of your domain vocabulary. Version B has every industry-specific noun removed and every assumption about the solution stripped out — a pure functional statement.

Run both through the same model. Compare the return sets side by side, and count how many results in each set come from outside your own sector.

Facilitator script

1. 5 min · Write version A
   Write the brief the way you would send it to a colleague. Jargon welcome. This is your control condition.
2. 8 min · Strip it
   Now remove every noun that names your industry, your product, or your existing solution. If someone outside your sector could not tell what business you are in, you have done it right. Keep the function, lose the furniture.
3. 10 min · Run both
   Same model, same settings, both briefs. Do not clean up the results yet.
4. 7 min · Count the outsiders
   How many results in each set come from outside your sector? That difference is the cost your own vocabulary has been quietly imposing on every search you have ever run.

The evidence

Hassan & Nair (2026), Research Policy, identify de-contextualizing as one of three AI complementarities in innovation search, derived inductively from 41 interviews across 27 organisations. Participants described AI as returning options they had already ruled out on assumed grounds — one R&D manager reported selecting two technologies that human search had dismissed on feasibility assumptions. Another noted that AI has no gut feeling, so it never pre-dismisses; a human "just neglects" an option and then "might miss something important."

The authors frame the human–AI pairing explicitly as divergent and convergent thinking, quoting one participant: AI helps you go broader more quickly, but at some point you have to converge — and AI is much less useful on that half.

Evidence profile

Framework-grade. Inductive interview study; no outcome was measured and no comparison was run.

Hassan, R., & Nair, S. (2026). Searching together: Human–AI complementarities in innovation search. Research Policy, 55, 105597. Open access, CC BY. Find this paper ↗

How to run it

Run the outsider seat protocol, but assign the role to the search tool. Give it the decontextualised brief from activity 20, and require it to return results from at least five named sectors none of which is yours.

Then — and this is the part teams skip — assign a human translator whose only job is to map each foreign result back to a mechanism in your domain. Without the translator you get a list of curiosities.

Facilitator script

1. 5 min · Name the forbidden sector
   Your own industry is off the table for this round. Anything the tool returns from inside our sector, we discard, however good it looks.
2. 15 min · Retrieve wide
   Five sectors minimum, none of them ours. Do not filter for relevance yet — filtering is the next phase and doing it now collapses the whole exercise.
3. 15 min · Translate, do not admire
   For each result, one sentence: what is the mechanism, and what is its equivalent here? Not "that is interesting" — a mechanism and a mapping.
4. 5 min · Debrief the discomfort
   Which results did you want to throw away fastest? Note them. That instinct is the thing this activity exists to interrupt.

The evidence

Hassan & Nair (2026) identify boundary-spanning search as AI's primary complementary capability — crossing regional, linguistic, disciplinary and organisational boundaries. Participants described AI search as "not based on experience, it is based on what is available," and noted it "doesn't care where it comes from" geographically.

The mechanism is the same one Jeppesen & Lakhani (2010) measured quantitatively across 166 InnoCentive challenges and 12,000+ solvers: the probability of a winning solution rose with the distance between the solver's expertise and the problem domain. That study measured outcomes; this one describes a machine doing the same job. Use the older paper for the evidence and the newer one for the practice.

Evidence profile

Framework-grade for the AI half; the underlying distance-and-solving effect (activity 09) is where the outcome evidence lives.

Hassan, R., & Nair, S. (2026). Research Policy, 55, 105597. Read alongside Jeppesen, L. B., & Lakhani, K. R. (2010), Organization Science, 21(5), 1016–1033. Find this paper ↗

How to run it

Do not delete the discard pile. Take the results your search tool returned that were clearly wrong, irrelevant, or fabricated, and run them through one question: if this were the answer, what would the question have been?

Most produce nothing. The exercise costs twenty minutes and occasionally produces a problem nobody had articulated.

Facilitator script

1. 5 min · Recover the discards
   Everything we threw out in the last hour — put it back on the table. Yes, including the ones that were obviously nonsense.
2. 15 min · Reverse the pair
   For each one: if this were a good answer, what problem would it be answering? Write the problem, not a critique of the answer.
3. 10 min · Test against reality
   Do any of those invented problems actually exist for us? Even faintly? Circle them.
4. 5 min · Set expectations honestly
   Most sessions produce nothing here and that is the expected result. This is a lottery ticket with a twenty-minute price, not a method.

The evidence

Hassan & Nair (2026) argue that AI hallucination, used deliberately, can seed serendipitous discovery — connecting it to Hippel & Von Krogh's (2016) concept of need–solution pairs, where a viable solution reveals a problem nobody had formulated. The strongest supporting cases in the paper are real and published: de novo protein design by deep network hallucination (Anishchenko et al., 2021, Nature), and the Robot Scientist Eve, which while screening triclosan against one malarial enzyme target unexpectedly identified a different one, DHFR (Bilsland et al., 2018, Scientific Reports).

Those are genuine discoveries. What does not yet exist is any evidence that deliberately harvesting hallucinations works as a repeatable workshop method for ordinary problems.

Evidence profile

The weakest card in the library. Individual examples are documented; the general claim has never been tested.

Hippel, E. von, & von Krogh, G. (2016). Identifying viable need–solution pairs. See also Anishchenko, I., et al. (2021), Nature, 600(7889), 547–552. Find this paper ↗

How to run it

Two rules, both structural.

One: review the full return set, never the pre-filtered shortlist. Whoever filtered it applied their own contextual assumptions, which is exactly what the AI search existed to bypass.

Two: score novelty blind to source. Strip every result of whether it came from AI or human search before anyone rates it. Reveal the source only after scoring.

Facilitator script

1. 5 min · Demand the long list
   I do not want the shortlist. I want everything the search returned, including what was cut and who cut it.
2. 12 min · Strip and shuffle
   Remove all provenance. Nobody rating these should know which came from the tool and which from our own network.
3. 13 min · Score novelty only
   Novelty first, on its own, by people who are not scoring feasibility. If you catch yourself thinking "we could never do that", you are on the wrong axis.
4. 10 min · Force the advance and reveal
   One high-novelty concept advances with a named advocate. Now reveal the sources — and notice whether the AI-sourced items clustered in what you rejected.

The evidence

Hassan & Nair (2026) position human judging as the corrective that filters AI output and mitigates AI bias. That is the paper's weakest link, because it is never tested against Mueller, Melwani & Goncalo (2012), who showed that under uncertainty people develop an implicit bias against creativity that impairs their ability to recognise a creative idea — while sincerely reporting that they want one. Novel outputs from unfamiliar domains with uncertain relevance is a precise description of the uncertainty condition that triggers it.

Second reason for structure: Doshi & Hauser (2024), Science Advances. In an online experiment, writers given GPT-4 story ideas produced work rated more creative, better written and more enjoyable — with the largest gains among less creative writers, effectively equalising evaluations. But the AI-assisted stories were measurably more similar to each other than the human-only stories. The authors describe it as a social dilemma: individually better off, collectively a narrower range of novel content.

Evidence profile

Framework-grade. Derived from interviews, with no measured effect on selection quality.

Mueller, J. S., Melwani, S., & Goncalo, J. A. (2012). Psychological Science, 23(1), 13–17. Doshi, A. R., & Hauser, O. P. (2024). Generative AI enhances individual creativity but reduces the collective diversity of novel content. Science Advances, 10(28), eadn5290. Find this paper ↗

How to run it

Split the room in two, out of earshot. Group A gets the design brief alone. Group B gets the brief plus one worked example, and — this is the part that makes the demonstration land — an explicit instruction that the example contains three named flaws which they must avoid.

Both groups ideate for the same time. Then post the outputs together and count how many of Group B's ideas carry each named flaw. Historically, most of them do.

Facilitator script

1. 3 min · Split and brief separately
   I am going to give the two halves of the room slightly different instructions. Do not compare notes until we post the results — the whole exercise depends on it.
2. 12 min · Ideate
   Twelve minutes, sketch or write as many distinct designs as you can. Group B, remember: the example has three flaws, listed at the bottom. Avoid them.
3. 10 min · Post and count
   Everything on the wall, Group A on the left, Group B on the right. Now we count. How many designs on the right contain flaw one? Flaw two? Flaw three?
4. 10 min · Debrief the mechanism
   Group B were warned. In writing. And it made almost no difference, because fixation is not a failure of attention — it is what happens when an example becomes the category you search inside. Now: what is the example that is already sitting in your team's head before every kickoff?

The evidence

Jansson & Smith (1991), Design Studies, gave engineering designers a brief with or without an example solution. Those shown the example reproduced its features — including features they had been explicitly told were defective — at markedly higher rates. The authors named the effect design fixation.

It is among the most robustly replicated results in design cognition. Purcell & Gero (1996) replicated it and showed it varies by discipline; Chrysikou & Weisberg (2005) replicated it in a psychology sample and showed that a defixating instruction can partly reduce it; Linsey et al. (2010) demonstrated it in expert faculty engineers, not just students, and tested countermeasures.

Two things make this stronger evidence than most of the library. The outcome is real design output rather than an Alternate Uses score, and the effect appears in experts, which is where most creativity findings quietly disappear.

Evidence profile

Controlled between-subjects experiment, replicated across four decades and multiple domains, scored on real design output.

Jansson, D. G., & Smith, S. M. (1991). Design fixation. Design Studies, 12(1), 3–11. doi:10.1016/0142-694X(91)90003-F Find this paper ↗

How to run it

This is the operational core of C–K design theory (see activity 26) reduced to something you can run in forty minutes.

Take one brief. Prepare two example sheets. The restrictive sheet shows a solution from inside the obvious path everyone already searches. The expansive sheet shows a solution from a path almost nobody considers. Assign the room at random — genuinely at random, count off — and give each half one sheet. Ideate, then score originality against the mapped solution space rather than by gut feeling.

The classic vehicle is the egg task: ensure a hen's egg dropped from ten metres does not break. Almost everyone searches the same region — wrap it, cushion the landing, slow the fall. The expansive paths are the ones that change the egg, change the ground, or change what "not break" means.

Facilitator script

1. 6 min · Map the solution space first
   Before we ideate, we map. On the board: what are the families of solution everyone reaches for? Cushion it, slow it, protect it. Those are the paths inside the obvious set. Anything not on this board is the expansive region.
2. 4 min · Randomise and distribute
   Count off, ones and twos. Ones take sheet A, twos take sheet B. Do not read your neighbour's sheet.
3. 12 min · Generate
   Twelve minutes. As many distinct solutions as you can. One per card.
4. 10 min · Score against the map
   Sort every card onto the map we drew. Anything landing outside the mapped families counts as expansive. Count the two groups separately.
5. 8 min · Reveal and generalise
   You had different sheets. Group A's example sat inside the obvious set; Group B's sat outside it. Look at the counts. The lesson is not that examples help or hurt — it is that the example silently tells people where the search space ends.

The evidence

Agogué, Kazakçi, Hatchuel, Le Masson, Weil, Poirel & Cassotti (2014), Journal of Creative Behavior, ran the egg task with restrictive and expansive example conditions. Restrictive examples — those lying within the commonly explored solution path — depressed originality relative to no example. Expansive examples — lying outside it — raised originality. The manipulation was the position of the example in the solution space, not its quality or detail.

The finding has an unusually good follow-up record for this literature. Agogué, Poirel, Pineau, Houdé & Cassotti (2014), Thinking Skills and Creativity, replicated the pattern across age groups and tested training. Agogué, Le Masson, Dalmasso, Houdé & Cassotti (2015), Psychology of Aesthetics, Creativity, and the Arts, found professional industrial designers and engineers resist classical solutions differently from novices. Ezzat et al. (2020) separated the specificity of an example from its abstraction and found opposing effects.

What raises this above a single cute result is that originality is scored against an independently mapped solution space rather than by rater impression, which removes a large chunk of the subjectivity that weakens most divergent-thinking scoring.

Evidence profile

Controlled experiment with random assignment, independently replicated in adults, children and professional designers; originality scored against a mapped solution space.

Agogué, M., Kazakçi, A., Hatchuel, A., Le Masson, P., Weil, B., Poirel, N., & Cassotti, M. (2014). The impact of type of examples on originality. The Journal of Creative Behavior, 48(1), 1–12. doi:10.1002/jocb.37 Find this paper ↗

How to run it

C–K theory, developed by Hatchuel and Weil at MINES Paris, models designing as the joint expansion of two spaces. K-space holds propositions with a known logical status — things you can call true or false. C-space holds undecidable propositions: "a boat that has no hull" is neither true nor false given current knowledge, which is precisely what makes it a concept rather than a fact.

Designing is the four operators between them: C→C (partitioning a concept into more specific ones), C→K (asking what knowledge a concept demands), K→C (using knowledge to open a new concept), K→K (ordinary learning). A design converges when a concept acquires a logical status and crosses into K.

Run it on the wall. Two colours: one for K, one for C. Start with the brief as a C-proposition. Every time someone says "but we know that…", that is a K card. Every time a K card makes a new undecidable proposition available, draw the arrow.

Facilitator script

1. 8 min · Write the initial concept
   State the brief as a proposition we cannot yet call true or false. Not 'improve the onboarding' — that is a goal. Something like 'an onboarding with no first session'. If someone can immediately tell me it is true or false, it is not a concept, it is knowledge.
2. 12 min · Harvest K
   What do we actually know that bears on this? Constraints, results, prior attempts, regulations. Blue cards. Include the things we know that we wish were not true.
3. 15 min · Partition the concept
   Now split the concept into narrower ones. Two kinds of split: restrictive, which adds a property from inside what we already do, and expansive, which adds a property that breaks an assumption in K. Mark every expansive partition with a star. Most teams produce almost none, and that is the finding.
4. 15 min · Follow the starred branches
   For each starred branch: what would we have to know for this to become decidable? That question is the C→K operator, and the answer is your actual research agenda.
5. 10 min · Close the loop
   Which new knowledge, once acquired, would open a concept we could not previously state? That is where a design programme comes from, as opposed to a list of ideas.

The evidence

Hatchuel & Weil (2009), Research in Engineering Design, give the formal statement of C–K theory: design as the co-expansion of a Concept space of undecidable propositions and a Knowledge space of decidable ones, connected by four operators. The theory's distinctive claim is that creativity is not a psychological trait but a logical operation — expansive partitioning — which can be taught and inspected.

The evidence has two very different halves, and the honest thing is to keep them apart.

The theory itself is framework-grade. Its support is industrial case studies and action research — KCP workshops at Renault, Thales, RATP and others, documented by Le Masson, Weil & Hatchuel. No randomised comparison against an active control has tested whether C–K sessions produce more or better innovation than another structured method. Nobody has measured the outcome. Teach the mechanism; do not quote a number.

Its fixation predictions are effect-grade. The expansive-versus-restrictive prediction has been tested experimentally and replicated — see activity 25. That is a genuine and unusual asset: most design methodologies in circulation have no experimental arm at all.

Evidence profile

Framework-grade. A formal design theory supported by industrial case studies and action research; the theory itself has not been tested against an active control for creative output.

Hatchuel, A., & Weil, B. (2009). C–K design theory: an advanced formulation. Research in Engineering Design, 19(4), 181–192. See also Le Masson, Weil & Hatchuel, Design Theory (Springer, 2017). doi:10.1007/s00163-008-0043-4 Find this paper ↗

How to run it

Run the same problem under two conditions with two groups. Condition A gets the standard rule: no criticism, defer judgment. Condition B gets an explicit instruction to debate and to challenge each other's ideas as they arrive.

Count total ideas and, separately, ideas generated afterwards when both groups are asked individually for anything else that occurs to them. The second count is where the interesting difference usually shows up.

Facilitator script

1. 5 min · Set the two rules
   Group A: the classic rule. Nothing is criticised, all ideas are welcome, judgment is deferred. Group B: debate as you go. If you think an idea is wrong, say so and say why. This is not permission to be unpleasant — it is an instruction to engage.
2. 20 min · Run both
   Twenty minutes on the same problem. Capture everything, including things that were argued down.
3. 10 min · The individual tail
   Now, alone and in silence: anything else that has occurred to you since. Five minutes, write it down. This is a separate count.
4. 10 min · Compare and complicate
   Two numbers per group: in-session ideas and tail ideas. Then the harder question — which room would you rather have been in, and is that the same as which room produced more?

The evidence

Nemeth, Personnaz, Personnaz & Goncalo (2004), European Journal of Social Psychology, compared brainstorming instructions with debate instructions in samples from the United States and France. The debate condition produced more ideas, and the advantage persisted into the individual ideas participants reported afterwards. The finding sits in a longer programme on minority dissent, in which exposure to a persistently argued minority view broadened rather than narrowed search — measured on word-association and problem-solving tasks, not on self-report.

This is a direct challenge to Osborn's defer-judgment rule, which activity 01 uses. Both cannot be straightforwardly right, and it would be dishonest for this library to carry one and quietly omit the other.

The reconciliation most defensible on current evidence: the harm in ordinary group brainstorming is production blocking and evaluation apprehension, not disagreement as such (see activity 07). Structured, task-focused dissent with everyone still able to produce appears to help; unstructured social evaluation does not. That reading is plausible and partly supported — it is not settled.

Evidence profile

Controlled laboratory experiments with random assignment, replicated across cultures; outcomes are validated ideation and association measures rather than field innovation.

Nemeth, C. J., Personnaz, B., Personnaz, M., & Goncalo, J. A. (2004). The liberating role of conflict in group creativity. European Journal of Social Psychology, 34(4), 365–374. doi:10.1002/ejsp.210 Find this paper ↗

How to run it

The mechanism is not "take a break." It is that sleep reorganises recently encoded material, and insight sometimes falls out of that reorganisation. Which means the scheduling instruction is specific: the problem has to be encoded before sleep, in enough depth that there is something to consolidate.

Day one: pose the problem, work it hard enough that everyone has hit a wall, then stop — explicitly, with no request to think about it overnight. Day two: first thing, individually and in silence, before any discussion.

Facilitator script

1. 10 min · Encode properly, day one
   Ten minutes on the problem, and I want you to get stuck. Being stuck is the condition, not the failure. Write down where the wall is.
2. 2 min · Close it down
   We stop here. I am deliberately not asking you to think about this tonight — the effect does not require deliberate rumination and asking for it tends to produce anxiety instead.
3. 8 min · Day two, silent first
   Before anyone speaks: eight minutes, alone, on the same problem. If something arrived overnight, it will not survive the first three minutes of group conversation, so write it before we talk.
4. 5 min · Collect the overnight set
   Hands up if something shifted. Now compare where the wall was yesterday to where it is now — the interesting cases are where the wall moved rather than where an answer appeared.

The evidence

Wagner, Gais, Haider, Verleger & Born (2004), Nature, trained participants on a number-reduction task containing a hidden shortcut rule. Participants who slept between training and retest discovered the rule at more than twice the rate of matched groups who stayed awake for equivalent periods, day or night. The comparison against a wake-control of the same duration is what makes it more than a rest effect.

The wider sleep-and-memory consolidation literature is strong and heavily replicated. The specific leap from consolidation to creative insight is the softer link — the number-reduction task is a rule-discovery task, and generalising it to open-ended ideation is an inference the study does not make.

Read alongside activity 06, which uses the very different sleep-onset (N1) window, and activity 04, whose incubation meta-analysis covers the waking case.

Evidence profile

Controlled experiment with matched wake-control groups; replicated in related paradigms, though the specific insight result has been contested on task-specificity grounds.

Wagner, U., Gais, S., Haider, H., Verleger, R., & Born, J. (2004). Sleep inspires insight. Nature, 427, 352–355. doi:10.1038/nature02223 Find this paper ↗

How to run it

The distinction is the whole activity. Focused-attention meditation trains sustained attention on a single object — breath, mantra, a point. Open-monitoring trains non-reactive awareness of whatever arises, without selecting.

Run both, ideally with the room split, and put an Alternate Uses Task afterwards. Then discuss why a room that assumes "mindfulness makes you creative" would have predicted no difference — and what it costs to be that imprecise.

Facilitator script

1. 4 min · Explain the split, not the prediction
   Two styles. One narrows attention onto a single object. One widens it and refuses to select. Do not tell each other which you think will help — I will ask afterwards.
2. 12 min · Run the practice
   Twelve minutes. Group A, attention on the breath; when it wanders, return it. Group B, notice whatever arises — sound, sensation, thought — and do not follow any of it, and do not push any of it away.
3. 8 min · Immediate AUT
   Straight into it, no discussion. Eight minutes: as many uses for a paperclip as you can. Quantity.
4. 6 min · Score and discuss
   Fluency counts first, then originality. Then the real question: if the two styles differ, what does that do to every claim you have heard that begins 'mindfulness improves…'?

The evidence

Colzato, Ozturk & Hommel (2012), Frontiers in Psychology, compared open-monitoring and focused-attention meditation within experienced practitioners and found open-monitoring improved divergent thinking on the Alternate Uses Task, while focused-attention did not — and if anything favoured convergent performance. Lippelt, Hommel & Colzato's (2014) review of the two styles reports a broadly consistent dissociation across studies.

The caution is real. Sample sizes in this literature are small, several studies use experienced meditators rather than naïve participants — which limits what a single workshop session can claim — and the broader meditation-and-cognition field has a well-documented publication-bias problem. Treat the dissociation as the finding worth teaching, not the magnitude.

Set against activity 16: this is better evidenced than the NSDR protocol on that card, and it is the version to run if you want a state intervention with an actual creativity outcome attached.

Evidence profile

Randomised laboratory comparison of two meditation styles against each other, replicated in independent samples; outcome is the Alternate Uses Task.

Colzato, L. S., Ozturk, A., & Hommel, B. (2012). Meditate to create: the impact of focused-attention and open-monitoring training on convergent and divergent thinking. Frontiers in Psychology, 3, 116. doi:10.3389/fpsyg.2012.00116 Find this paper ↗

How to run it

Give each participant three heuristic cards drawn at random — add motion, change the surface, make it hollow, use a common base, and so on. Each card is a single transformation with a worked visual example.

One sketch per card, then swap decks and repeat. The rule that makes it work: the heuristic is not optional. If the card says change the surface, the sketch must change the surface, however unpromising that seems.

Facilitator script

1. 5 min · Frame the deck honestly
   These are not somebody's clever acronym. They were derived by analysing what expert designers and award-winning products actually did, then reduced to the transformations that recurred. That provenance is why we are using this deck rather than a more famous one.
2. 15 min · First pass, three cards
   Three cards each, one sketch per card, five minutes per sketch. You must apply the card even if you think it is wrong for this problem. Bad forced sketches are the point — they move you out of your first solution.
3. 15 min · Swap and repeat
   Pass your cards two seats left. Three more sketches. Notice whether the second set is easier or harder than the first.
4. 10 min · Sort by distance
   Lay everything out. Which sketches are variations on where you started, and which are somewhere genuinely else? Count the second group per person. That count is what the deck is for.

The evidence

Design Heuristics were derived empirically by Yilmaz, Seifert, Daly and Gonzalez from protocol studies of practising designers, analysis of award-winning products and expert interviews — rather than being invented and then promoted. The resulting set of 77 heuristics has been tested in controlled studies with engineering and industrial design students, with concepts rated by independent judges for novelty and quality; heuristic-supported groups produced more varied and more novel concepts than unsupported controls.

Two reasons this earns a place over the better-known idea-prompt methods. The outcome measured is rated design output rather than an Alternate Uses score. And the tool's content has a documented derivation, so a facilitator can answer "where did these prompts come from" with something other than "a book."

The limits are ordinary but worth stating: student samples predominate, judge-rated novelty carries the usual subjectivity, and the comparison is often against no support rather than against another prompt method — so the deck beating nothing is better established than the deck beating SCAMPER.

Evidence profile

Controlled studies with random assignment in engineering and industrial design samples; outcomes are rated design concepts, not lab proxies.

Yilmaz, S., Daly, S. R., Seifert, C. M., & Gonzalez, R. (2016). Evidence-based design heuristics for idea generation. Design Studies, 46, 95–124. doi:10.1016/j.destud.2016.05.001 Find this paper ↗

How to run it

Take a genuine shortlist — ideas you have not yet decided between. Have three groups forecast success independently and in private: the people who created the ideas, the people who manage the portfolio, and if you can get them, people from the intended audience.

Do not average them into a single number. Compare the shape of the three rankings, and specifically look at which ideas the manager group ranks lowest that another group ranks high. Those are the ideas your normal process kills.

Facilitator script

1. 5 min · Assign roles honestly
   Three sheets. You are scoring as creator, as manager, or as audience. If you happen to be both a creator and a manager here, score as creator and hand your manager hat to someone else — mixing the two is what we are trying to see.
2. 12 min · Private forecasting
   Score every idea for likely success, alone, no discussion. Private is not a formality — one confident voice collapses the whole comparison.
3. 13 min · Three rankings, side by side
   Post all three. Do not average. I want to see the disagreements, because the disagreements are the finding.
4. 10 min · Find the killed ideas
   Which ideas do managers rank in the bottom third and creators or audience rank in the top third? Name them. In your normal process, every one of those is dead by Thursday.

The evidence

Berg (2016), Administrative Science Quarterly, had creators, managers and audience members forecast the success of novel ideas, then compared those forecasts against real downstream outcomes. Creators were better forecasters of novel ideas than managers on the measures used — the reverse of the assumption embedded in most stage-gate processes. The proposed mechanism is a divergent-then-convergent asymmetry: the same mindset that generates a novel idea also supports evaluating novelty, while a purely evaluative mindset systematically discounts it.

Read with activity 11. Mueller, Melwani & Goncalo's bias against creativity explains why the evaluative role penalises novelty; Berg's work says something about who to ask instead.

The evidence limit is clear. The role comparison is observational — people were not randomly assigned to be creators or managers — so selection is a live alternative explanation, and the finding has not been independently replicated. What earns its place is that the outcome is real success data rather than a lab proxy.

Evidence profile

Field study with real innovation outcomes and pre-registered-style forecasting design, but observational role comparison rather than random assignment; not yet independently replicated.

Berg, J. M. (2016). Balancing on the creative highwire: forecasting the success of novel ideas in organizations. Administrative Science Quarterly, 61(3), 433–468. doi:10.1177/0001839216642211 Find this paper ↗

How to run it

Run a short structured improv sequence — word association at speed, "yes, and" building, one-word-at-a-time storytelling — with the group standing, and with the explicit rule that the aim is not to be funny.

What makes this an evidence-informed activity rather than an icebreaker is the debrief. Improv's studied benefit is not team bonding, which is what it is usually sold as. It is a measured change in divergent thinking and in tolerance for uncertainty. Say that, and measure it.

Facilitator script

1. 5 min · Set the anti-comedy rule
   The aim is not to be funny. If you are trying to be funny, you are editing, and editing is the thing this exercise exists to suppress. Speed and acceptance, nothing else.
2. 10 min · Word association at speed
   Circle, one word each, no pauses. If you hesitate more than two seconds we start again. Nonsense is fine. Repetition is fine.
3. 15 min · Yes, and
   In pairs, build a scenario. Every line begins by accepting what came before and adding one thing. You may not block, and you may not ask a question that hands the work back.
4. 10 min · One word at a time
   Whole group, one story, one word each around the circle. Notice the moment you want to steer it somewhere — and let it go.
5. 10 min · Debrief the real claim
   What this is evidenced for is divergent thinking and comfort with uncertainty. It is not evidenced as a team-building intervention, whatever the person who sold it to your company said. What actually changed in the last forty minutes?

The evidence

Felsman, Gunawardena & Seifert (2020), in Thinking Skills and Creativity and related work, compared short improvisation training against an active control activity and found gains in divergent thinking and in tolerance of uncertainty, with affect improvements alongside. The use of an active control rather than a do-nothing group puts this above most of the training literature Sio & Lortie-Forgues criticised.

The limits are the usual ones for this corner of the field: modest samples, mostly student participants, short follow-up, and little independent replication. The direction is consistent; the magnitude is not established.

Note what the studies do not show. The common corporate justification — improv builds trust and collaboration — is not what was measured. If you want the climate effect, activity 10 has far better evidence for it.

Evidence profile

Randomised comparison against an active control condition, with divergent-thinking and affect outcomes; small samples and limited independent replication.

Felsman, P., Gunawardena, S., & Seifert, C. M. (2020). Improv experience promotes divergent thinking, uncertainty tolerance, and affective well-being. Thinking Skills and Creativity, 35, 100632. doi:10.1016/j.tsc.2020.100632 Find this paper ↗

How to run it

Most "get the room in a good mood" advice is wrong in a specific and correctable way. What the meta-analytic evidence supports is not positive mood but activating positive mood — high-arousal states such as elation or excitement. Low-arousal positive states like calm and relaxation show little effect on idea generation.

Run two short inductions and a divergent task after each, then have the room plot themselves on a two-by-two: pleasant/unpleasant against activated/deactivated. The two-by-two is what people take away.

Facilitator script

1. 5 min · Draw the two-by-two
   Two axes, not one. Pleasant to unpleasant across, activated to deactivated up. Four quadrants. Most advice about mood and creativity only knows about the left-right axis, and that is why it under-performs.
2. 8 min · Induction one, then task
   Short high-energy piece, then four minutes of divergent generation immediately. Do not discuss between them.
3. 8 min · Induction two, then task
   Now something calm and pleasant, then the same four minutes on a different prompt.
4. 4 min · Plot and count
   Mark yourself in a quadrant for each round and count your outputs. This is a demonstration, not an experiment — order effects alone could produce the difference. But the quadrant map is the durable part.

The evidence

Baas, De Dreu & Nijstad (2008), Psychological Bulletin, meta-analysed 25 years of mood-and-creativity research. The headline is the qualification: positive mood benefits creativity when it is activating; deactivating positive moods show little or no benefit. Activating negative moods such as anger can also raise output, though through a different route — persistence rather than flexibility — and at a cost worth thinking about before you engineer it.

This connects to the dual-pathway account (Nijstad, De Dreu, Rietzschel & Baas, 2010): creative output can be reached either by flexibility, moving between categories, or by persistence, going deep within them. Different states favour different routes, which is why a single "be positive" instruction under-performs.

Grade honestly. This is a large, well-conducted meta-analysis, and it is also mood-induction studies with lab divergent-thinking outcomes, mostly short-term, mostly students. And by Sio & Lortie-Forgues's own moderator analysis, motivational and affective routes underperform cognitive-skills training — so this is a supporting card, not a headline one.

Evidence profile

Large meta-analysis of mood-induction experiments; outcome measures are predominantly lab divergent-thinking tasks.

Baas, M., De Dreu, C. K. W., & Nijstad, B. A. (2008). A meta-analysis of 25 years of mood-creativity research. Psychological Bulletin, 134(6), 779–806. doi:10.1037/a0012815 Find this paper ↗

How to run it

SCAMPER — Substitute, Combine, Adapt, Modify, Put to another use, Eliminate, Reverse — was assembled by Bob Eberle from Osborn's checklist questions. It is a mnemonic over a list, not a research finding, and it has never been the subject of anything resembling the trials this site asks for elsewhere.

So run it twice. First as a technique, honestly, because it does generate ideas. Then as a source-tracing exercise: ask the room to find the study. Fifteen minutes with a search engine is usually enough for the point to land on its own.

Facilitator script

1. 15 min · Run it straight
   Seven prompts, one pass each. Substitute what? Combine with what? Keep going through the list. Capture everything — this half is not a trick, the technique does produce output.
2. 15 min · Now find the evidence
   Search for the controlled trial. Randomised, active control, creativity outcome. You have fifteen minutes. Tell me what you find and, more importantly, what the studies you do find actually compared against.
3. 10 min · Separate the two questions
   Two different questions, and rooms collapse them constantly. Does this produce ideas? Almost certainly yes — any structured prompt beats staring at a wall. Does it produce more or better ideas than another structured prompt? Nobody has really tested it. Only the second question justifies teaching it as a method.

The evidence

SCAMPER's lineage runs from Osborn's Applied Imagination (1953) checklist questions to Eberle's mnemonic (1971). Neither is a controlled study. The published evaluation literature consists mainly of small quasi-experimental studies, frequently with school-age participants, frequently comparing a SCAMPER-based programme against no treatment, and almost never isolating SCAMPER from the surrounding instruction, facilitation and attention.

By the standards Sio & Lortie-Forgues apply — randomisation, an active control, a pretest — essentially none of it qualifies. This does not mean SCAMPER does not work. It means the claim has not been tested, which is a different statement, and the difference is what this whole site is about.

Compare activity 30. Design Heuristics prompts overlap substantially with SCAMPER prompts in content; the difference in grade is entirely a difference in who did the studies.

Evidence profile

Widely used, weakly tested. Small quasi-experimental studies, mostly with children, mostly against no-treatment controls, rarely isolating SCAMPER from the training that surrounds it.

Eberle, B. (1971). SCAMPER: Games for Imagination Development. Buffalo, NY: D.O.K. Publishers. Derived from Osborn, A. F. (1953), Applied Imagination. Find this paper ↗

How to run it

De Bono's six hats assign the whole group a single mode at a time: white for information, red for feeling, black for caution, yellow for benefit, green for new ideas, blue for process. Run it on a real decision so people can feel what it does.

Then separate the framework from its one defensible mechanism. The instruction everyone does the same kind of thinking at the same time is a form of generate/evaluate separation, and that has independent support — see activity 13. The rest of the apparatus — six specific modes, these colours, this sequence — does not.

Facilitator script

1. 5 min · Explain parallel thinking
   The rule is that we are all in the same mode at once. Nobody is the designated sceptic and nobody is the designated enthusiast. That single rule is the part worth keeping.
2. 22 min · Run the sequence
   White first — only what we know, no interpretation. Then green, only new possibilities. Then yellow, then black, then red. Blue is mine. Roughly four minutes each, and stay in the mode.
3. 8 min · Isolate the active ingredient
   Which part of the last twenty minutes did the work? Be specific. My claim is that it was the separation, not the colours — argue with me.
4. 10 min · Check the evidence chain
   De Bono's method has sold in the millions and been adopted by governments. Search for the randomised trial with an active control. Then ask what commercial success is actually evidence of.

The evidence

De Bono's Six Thinking Hats (1985) is a practitioner framework. Its published evaluation base is dominated by case reports, organisational testimonials, satisfaction surveys and small studies without active controls. There is no body of randomised trials establishing that the six-mode structure outperforms another structured discussion protocol on creative output.

What can be defended is narrower. Separating generation from evaluation has good support (activity 13), including a mechanistic account in the network-switching literature in section 03. Reducing evaluation apprehension in groups has support (activities 07 and 10). The hats are one implementation of principles evidenced independently of the hats.

Teach it that way and you have kept everything usable while dropping the unsupported claim. That is a better outcome than either uncritical adoption or dismissal.

Evidence profile

Very widely adopted, very thinly evidenced. Mostly case reports and satisfaction surveys; the one defensible component is separation of thinking modes, which is evidenced elsewhere.

de Bono, E. (1985). Six Thinking Hats. Boston: Little, Brown. Read against Nijstad & Stroebe (2006) on group ideation mechanisms. Find this paper ↗

How to run it

Empathise, define, ideate, prototype, test. Put the diagram up and ask the room to grade each stage separately against the criteria in section 05 — what was measured, in whom, by whom.

The exercise works because the answers genuinely differ by stage. Prototyping and iterative testing have real support in product development research. Problem definition maps onto the problem-construction literature in activity 08, which is decently evidenced. The claim that these five stages in this order constitute a validated process is a different claim, and it has not been tested against any alternative ordering.

Facilitator script

1. 5 min · Put the diagram up
   Five boxes, five arrows. Almost everyone in this room has seen it. Today we are grading it box by box rather than accepting or rejecting the whole thing.
2. 18 min · Grade each stage
   For each stage: what is the primary evidence that this stage, as specified, improves outcomes? Not whether it feels right. What was measured, in whom.
3. 9 min · Find the untested claim
   Now the sequence itself. Where is the study comparing this order against another order? If we cannot find one, the arrows are a pedagogical convenience that got promoted.
4. 8 min · Keep what survives
   What would you still run tomorrow, and what would you stop calling evidence-based? That is a shorter list and a more honest one, and your students will trust it more.

The evidence

The five-stage formulation popularised by the Stanford d.school and IDEO is a teaching model, not a research finding. The surrounding literature is dominated by case descriptions and practitioner accounts; systematic reviews of design thinking's effectiveness repeatedly report heterogeneous definitions, weak measurement and few controlled comparisons.

Components fare differently. Iterative prototyping and early user testing have support in new product development research. Problem construction has support (activity 08). Empathy work as specified — short-form interviews producing personas — has notably little controlled evidence for its effect on creative output, whatever its other merits.

The interesting failure mode is the sampling frame from section 05. Much of the visible evidence for design thinking comes from consultancies and schools that teach it, which is not disqualifying but does mean the enthusiastic literature and the independent literature are not the same literature.

Evidence profile

Framework-grade with a large case literature and a small, mixed outcome literature; the specific five-stage sequence has not been tested against alternative structures.

Brown, T. (2008). Design thinking. Harvard Business Review, 86(6), 84–92. Read against systematic reviews of design thinking effectiveness, e.g. Micheli et al. (2019), Journal of Product Innovation Management, 36(2), 124–148. doi:10.1111/jpim.12466 Find this paper ↗

How to run it

TRIZ's core move is worth teaching regardless of its evidence grade: state the problem as a contradiction — improving A degrades B — rather than as a goal. Then look up the contradiction in the matrix and read off the inventive principles that historically resolved that pairing.

Run it on a real technical problem. The forced move from "make it better" to "these two parameters are in conflict" is where most of the value sits, and it survives even if the matrix does not.

Facilitator script

1. 10 min · State the contradiction
   Not a goal. A conflict. Something of the form: if we increase this, that gets worse. If you cannot write it in that form, you have not analysed the problem yet.
2. 8 min · Locate it in the matrix
   Find your two parameters, read the cell, note the principles it suggests. Do not evaluate them yet.
3. 17 min · Apply the principles
   For each suggested principle, force an application. Segmentation — what would it mean to divide this? Prior action — what could be done in advance? Force it even where it seems absurd.
4. 15 min · Audit the derivation
   Now the source question. Altshuller derived these by analysing patents. What kind of claim does that support, and what kind does it not? Specifically: a patent database tells you what was invented and granted. Does it tell you what works, or what gets filed?

The evidence

Altshuller derived TRIZ's 40 inventive principles and the contradiction matrix from analysis of large volumes of patents, identifying recurring resolution patterns. As a derivation this is more interesting than pure invention — it is at least grounded in a real corpus.

The evidential problem is what that corpus can support. Patent analysis is a study of what was filed and granted, which is not the same as what solved problems well, and it is silent on whether a person given the matrix outperforms a person given another structured method. The applied literature is dominated by industrial case studies and success reports, with very few controlled comparisons and almost no randomisation.

The contradiction formulation itself sits closer to defensible ground: it overlaps substantially with problem construction (activity 08) and with C–K's use of a knowledge constraint to open a concept (activity 26), both of which have better support.

Evidence profile

Derived from large-scale patent analysis rather than experiment; adoption evidence is industrial case studies, with almost no controlled comparison against other structured methods.

Altshuller, G. S. (1984). Creativity as an Exact Science. Gordon & Breach. See also Ilevbare, Probert & Phaal (2013), Technovation, 33(2–3), 30–37, on adoption evidence. doi:10.1016/j.technovation.2012.11.003 Find this paper ↗

How to run it

Do not run these as interventions. Run the trace, as with activity 19.

Hand out the three abstracts — blue versus red (Mehta & Zhu, 2009, Science), ceiling height (Meyers-Levy & Zhu, 2007), ambient noise at around 70 dB (Mehta, Zhu & Cheema, 2012). All three are published in strong outlets, all three are widely repeated in workplace design advice, and all three sit in the part of social and consumer psychology hit hardest by the replication crisis.

Ask the room how many of the three they had heard as settled fact before today.

Facilitator script

1. 4 min · Poll first
   Hands up: who has heard that blue rooms help creativity? High ceilings? Moderate background noise? Keep your hands up so you can see how much of the room shares each belief.
2. 12 min · Read the originals
   Three abstracts. For each: what was the sample, what was the outcome measure, and how large was the effect? Write the numbers down before you form a view.
3. 8 min · Read what happened next
   Now the replication record. Note specifically whether the follow-ups were larger, preregistered, and by independent groups — and what happened to the effect when they were.
4. 6 min · Reconcile with activity 14
   Here is a live contradiction inside this library. Activity 14 tells you to lower stimulation during generation. The 70-decibel finding says moderate noise helps. Both cannot be simply true. What would you have to know to decide?

The evidence

Colour. Mehta & Zhu (2009), Science, reported blue backgrounds favouring creative tasks and red favouring detail-oriented tasks. The effect has proved fragile: subsequent larger and preregistered attempts have found much smaller or null effects, and colour-cognition effects generally are among the more contested in the field.

Ceiling height. Meyers-Levy & Zhu (2007), Journal of Consumer Research, reported higher ceilings promoting abstract, relational processing. The construct-priming literature this belongs to has been badly damaged by replication failures across the board.

Ambient noise. Mehta, Zhu & Cheema (2012), Journal of Consumer Research, reported moderate ambient noise (around 70 dB) improving creative cognition relative to low or high noise. Replication attempts have been mixed, with several failing to recover the inverted-U pattern.

And the honest consequence for this library: activity 14 is not as secure as its moderate grade implies. The alpha-synchronisation and eye-closure evidence behind it is reasonable, but it points in the opposite direction to the noise finding, and this site should not carry one and hide the other. The current reading — that the two studies measure different things, internally-directed generation versus a broad creative-cognition composite — is a reconciliation, not a result.

Evidence profile

Well-designed original experiments in good journals whose effects have weakened or failed under larger, preregistered replication attempts.

Mehta, R., & Zhu, R. (2009). Blue or red? Science, 323(5918), 1226–1229. Mehta, Zhu & Cheema (2012), Journal of Consumer Research, 39(4), 784–799. Meyers-Levy & Zhu (2007), Journal of Consumer Research, 34(2), 174–186. doi:10.1126/science.1169144 Find this paper ↗

How to run it

Leung and colleagues (2012) reported that embodying metaphors improved creativity — including participants who literally sat outside a five-foot cardboard box performing better on creativity measures than those inside it.

Run this exactly as activity 19: as a trace, not a technique. It is arguably the best single teaching case in the library, because the original is genuinely clever and nothing about it looks careless.

Facilitator script

1. 4 min · Tell it as a fact
   Here is a finding. People who sat outside a physical box scored higher on creativity tasks than people who sat inside one. Published in a leading journal. How plausible does that sound?
2. 8 min · Read the original
   The design is good. Multiple experiments, several metaphors, sensible controls. Note what impresses you — that reaction is the thing we are examining.
3. 8 min · Read the replication record
   Now the follow-ups. Larger samples, preregistration, independent labs. What happened to the effect?
4. 5 min · Extract the rule
   Nobody here did anything wrong. The original was not sloppy and the replicators were not hostile. So what is the rule you take away? Mine is: a single elegant study is a hypothesis, and elegance is not evidence.

The evidence

Leung, Kim, Polman, Ong, Qiu, Goncalo & Sanchez-Burks (2012), Psychological Science, reported a set of embodied-cognition effects on creativity, including the literal box manipulation. The paper was widely covered and became a standard illustration in creativity teaching.

It belongs to the embodied-priming literature that the replication crisis affected most severely. Larger and preregistered replication efforts across this family of effects have generally failed to recover them at the reported magnitudes, and the field's confidence in short-manipulation priming of complex cognition has fallen substantially since 2012.

What makes it valuable teaching material is that the failure is not attributable to misconduct or incompetence. It is a well-executed study of an effect that appears not to be there, or to be far smaller than a small sample could detect. That is the ordinary case, and facilitators should expect it rather than treat it as scandal.

Evidence profile

Original controlled experiments in a leading journal; subsequent larger and preregistered replications did not recover the effect.

Leung, A. K.-y., Kim, S., Polman, E., Ong, L. S., Qiu, L., Goncalo, J. A., & Sanchez-Burks, J. (2012). Embodied metaphors and creative acts. Psychological Science, 23(5), 502–509. doi:10.1177/0956797611429801 Find this paper ↗

How to run it

This is a discussion card. Do not run it. It is in the library because it comes up in every workshop that touches creativity and drinking, and a facilitator should be able to say precisely what the study did and did not show.

Jarosz, Colflesh & Wiley (2012) brought participants to roughly 0.075 blood alcohol content and tested them on the Remote Associates Test. The intoxicated group solved more problems, and faster, and were more likely to report the solution as arriving suddenly rather than analytically.

The teaching value is in the boundary conditions, which almost nobody who repeats this finding can state.

Facilitator script

1. 4 min · State the finding precisely
   One specific blood alcohol level. One specific task — the Remote Associates Test, which measures convergent insight, not idea generation. One small sample. Say all three of those together or you have misreported it.
2. 7 min · Find the boundary
   What would you predict for divergent generation? For sustained work? For the same people two hours later? The study speaks to none of those, and the popular version speaks to all of them.
3. 9 min · Ask why it travelled
   This study is cited far more often in workshops than any of the meta-analyses on this site. Why? And what does that tell you about how findings are selected for transmission in your own field?

The evidence

Jarosz, Colflesh & Wiley (2012), Consciousness and Cognition, tested moderately intoxicated participants (around 0.075 BAC) against sober controls on the Remote Associates Test. The intoxicated group solved more items in less time and were more likely to characterise solutions as insight-like. The proposed mechanism is reduced executive control, broadening the search — the same mechanism proposed for the chronotype-mismatch effect in activity 15.

The limits are substantial. Small sample, a single task measuring convergent insight, no divergent measure, no assessment of real creative work, and no large replication. There is also a well-known asymmetry: the same reduced control that may help a three-word association task is very likely to hurt the sustained evaluative work that most creative production requires.

Included because it is a boundary case with a clear mechanism, and because being able to state a finding's limits precisely is the skill this library is actually trying to build.

Evidence profile

Small controlled laboratory experiment with an intoxicated and sober comparison on a convergent insight measure; not replicated at scale, and not a runnable workshop activity.

Jarosz, A. F., Colflesh, G. J. H., & Wiley, J. (2012). Uncorking the muse: alcohol intoxication facilitates creative problem solving. Consciousness and Cognition, 21(1), 487–493. doi:10.1016/j.concog.2012.01.002 Find this paper ↗