Social Cognition
Take or Pass Game
Also called: centipede game, backward induction game
Measure how far people let a pot grow: the participant and a simulated partner take turns to take the pot and end the game or pass it on, scored by the passes per game.
4-7 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Try it here first, with no account.
- Papers, 2021 to 2025
- 34
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures how far people pass a growing pot before taking it. In each game the participant and a simulated partner take turns, 6 at most, to take the pot and end the game or to pass it on so that it grows. The main score is how many of their 3 turns the participant passes, with games the partner ended early counted only as far as they went.
Constructs
- Reciprocal cooperation
- Trust in a partner's next move
- Reasoning ahead through turns
- Learning from repeated play
Research uses
- Behavioral economics research on trust and cooperation
- Studies of strategic reasoning and backward induction
- Research on social preferences
- Group comparison within a study
How the paradigm works
The Take or Pass Game measures how far people let a growing pot run before taking it. Two players take turns: the player whose turn it is can take the larger share of the pot, which ends the game, or pass, which makes the pot larger and hands the turn over. Reasoning back from the last turn, each player should take one turn before the other would, so standard game theory says the first player takes at once. People rarely do. In the first experiment, only 7% of four-move games and 1% of six-move games ended on the first move, the chance of taking rose toward the last move, and games ended earlier with experience (McKelvey and Palfrey, 1992). The game comes from Rosenthal (1981). A systematic review finds that people cooperate in it often, and that the payoffs, above all the gain from passing and its risk, change how much (Krockow, Colman and Pulford, 2016). Studies of chess players disagree: 69% of highly ranked players stopped at once in one (Palacios-Huerta and Volij, 2009), and almost none did in another, whose authors call the game ill-suited to testing backward induction (Levitt, List and Sadoff, 2011).
More about the paradigm
Here the participant moves first against a simulated partner, a new one in each game, with three turns each. The pot starts at 100 points and grows by 100 with each pass. Whoever takes it gets 70% and the other player 30%, and if the partner passes on the last turn the pot grows to 700 and the participant gets 70%. A table of what each player gets on every turn stays on screen, with the current turn marked, the way Crosetto and Mantovani (2018) showed the game; players stopped later when they had to work the payoffs out. The points are ConductScience's own: a linear game, because its rule is easy to convey. Each partner's turn to take is set before the session: of every 12 games, on its first turn in 1, its second in 4 and its third in 5, and never in 2, in an order drawn from the session's seed. So the chance of a take rises through the game, as in McKelvey and Palfrey's data, and nothing the participant does changes it. The standard instructions say each partner is a computer program that follows set rules; a neutral wording leaves that line out, and the last screen then says so. Studies usually explain the purpose afterwards. The standard session has 12 games (6 to 24). One check question with feedback comes first, there is no time limit, and the study decides what the points are worth.
The main score is how far the participant passes, in passes per game from 0 to 3. It is read from the take rate at each of the participant's three turns, so a game the partner ended counts only for the turns it reached and the partner's takes do not lower the score. Someone who always takes on their third turn scores 2, however often a partner took first. The take rates themselves are also reported, as are the score for the first and second half of the games and the change between them. Passing pays only if the partner passes too: with a take planned for the next turn, a pass pays more when the partner passes with a chance above about 7% on turn 1, 39% on turn 3 and 55% on turn 5. Against these partners, taking at once earns 70 points a game on average and passing twice about 249, the most of any fixed plan. A pass can reflect trust, kindness, a short horizon or a guess about a program, so the score describes choices in a game played for points, never a person's trust or reasoning, and it does not diagnose. People who know the partner is a program may pass differently from people playing a person, so compare people who saw the same instructions.
Scores
- Passes per Gamecount
- How far the participant passes, from 0 to 3, read from the take rate at each of their three turns, so that a game the partner ended counts only for the turns it reached.
- Read across a group; no better direction
- Take Rate, First Turn%
- Share of games in which the participant took the pot on their first turn, the game's equilibrium move.
- Informational
- Take Rate, Second Turn%
- Share of takes among the games that reached the participant's second turn; empty if none did.
- Informational
- Take Rate, Third Turn%
- Share of takes among the games that reached the participant's third turn; empty if none did.
- Informational
- Passes per Game, First Halfcount
- Passes per game over the first half of the games.
- Informational
- Passes per Game, Second Halfcount
- Passes per game over the second half of the games.
- Informational
- Change in Passes per Gamecount
- Passes per game in the second half minus the first.
- Below 0 means taking sooner with experience
- Points Wonpoints
- Points over all games.
- Informational
- Check Correctflag
- 1 when the check question on what a take pays was answered correctly, 0 when not.
- 1 supports interpretation
- Median Choice Timems
- Median time from a turn's screen to the choice, over choices at or after 100 ms.
- Informational
- Choices Under 100 mscount
- Choices made less than 100 ms after the turn's screen appeared.
- 0 expected
- Gamescount
- Games finished.
- Informational
Use and evidence
34 papers from 2021 to 2025 name this task in their title or abstract, down 21% compared with 2016 to 2020. That ranks it 343 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 11 |
| 2024 | 6 |
| 2023 | 4 |
| 2022 | 8 |
| 2021 | 5 |
| 2020 | 4 |
| 2019 | 8 |
| 2018 | 13 |
| 2017 | 7 |
| 2016 | 11 |
| 2015 | 12 |
| 2014 | 4 |
| 2013 | 11 |
| 2012 | 11 |
| 2011 | 6 |
| 2010 | 8 |
| 2009 | 10 |
| 2008 | 8 |
| 2007 | 7 |
| 2006 | 3 |
| 2005 | 6 |
| 2004 | 7 |
| 2003 | 4 |
| 2002 | 4 |
| 2001 | 2 |
| 2000 | 3 |
Use grade D
30 to 99 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 0
- Papers on reliability or validity
- 2
- Registered clinical studies using it as an outcome
- 0
Note on the counts: includes game-theory papers that run no task.
How the grades work
Use: A is 1,000 or more papers from 2021 to 2025, B 300 to 999, C 100 to 299, D 30 to 99 and E fewer than 30. Evidence: A means at least 25 papers report its test-retest reliability and at least 25 registered clinical studies use it as an outcome; B, at least 8 test-retest papers, and either 5 registered clinical studies or 150 reliability or validity papers; C, at least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies; D, little published measurement evidence found. The counts measure use. They include every paper that names the task in its title or abstract. Grades describe the published research on this task, not ConductCognition's version of it.
Source: ConductCognition test census 0.22.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Each turn
- The participant moves first. Whoever takes the pot gets 70% of it, the other player gets 30% and the game ends. A pass adds to the pot and hands the turn over. Each player has 3 turns, 6 in all.
- Points
- The pot starts at 100 points and grows by 100 with each pass. If the partner passes on the last turn, the pot grows to 700 and the participant gets 70%. A table of what each player gets on every turn stays on screen, with the current turn marked.
- Partner
- A computer program, a new one in each game. Of every 12 games it takes on its first turn in 1, on its second in 4 and on its third in 5, and passes all the way in 2, in an order drawn from the session's seed and saved. Nothing the participant does changes it. The standard instructions say each partner is a computer program that follows set rules; a neutral wording leaves that line out, and the last screen then says so.
- Games
- 12 (standard; 6 to 24), each followed by its result and the points so far. Points are hypothetical by default; a setting changes the instruction line to say the study pays.
- Check
- One question with feedback before the first game: the partner takes on turn 4.
- Scores
- Passes per game, from 0 to 3, read from the take rate at each of the participant's turns, so a game the partner ended counts only for the turns it reached; the three take rates; the same score over the first and second half of the games, and the change between them.
McKelvey and Palfrey (1992) matched students with one another for money, with piles that doubled. Here the partner is a computer program, the points are ConductScience's own and the pot grows in equal steps, because that rule is easy to convey (Crosetto and Mantovani, 2018). Passing pays only if the partner passes too: with a take planned for the next turn, a pass pays more when the partner passes with a chance above about 7% on turn 1, 39% on turn 3 and 55% on turn 5. People who know the partner is a program may pass differently from people playing a person, so compare people who saw the same instructions. Studies usually explain the purpose afterwards. The scores describe choices in a game played for points, never a person's trust or reasoning, and do not diagnose.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: the scored run's choices, one take or pass per turn. No choice is right or wrong, so accuracy is not judged against chance, and passing or taking on every turn is accepted. A run short of its planned games and a wrong check answer are flagged for review. Choices under 200 ms on at least half the turns invalidate.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. How far to pass depends on what a person expects of the partner, and McKelvey and Palfrey (1992) saw people pass at every chance and one take at every chance, so the same choice throughout is a real way to play. An answer under 200 ms is faster than a simple visual reaction, so it cannot reflect the item. Following response-time effort screening for untimed tests (Wise and Kong 2005), the session is invalid only when at least half the answers are that fast.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Games | 12 | 6 to 24 |
| Partner wording | A computer program (standard) | A computer program (standard), Neutral: your partner |
| Payment line | Hypothetical points (standard) | Hypothetical points (standard), The study pays for points |
Yearly plans can change settings within these limits. Free and pay-as-you-go studies use the standard settings.
Need a setting this test does not have?
Devices and languages
- Untimed test: runs on desktops, laptops, tablets and phones. Participant devices
- Languages: English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese.
- Available on: Every plan, including the free plan.
- No percentile. Scores and quality flags only. How norms work
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Take or Pass Game in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Rosenthal RW, 1981; McKelvey RD, Palfrey TR, 1992), with the standard settings (games: 12; partner wording: A computer program; payment line: Hypothetical points). The main score was passes per game (how far the participant passes, from 0 to 3, read from the take rate at each of their three turns, so that a game the partner ended counts only for the turns it reached).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Rosenthal, R. W. (1981). Games of perfect information, predatory pricing and the chain-store paradox. Journal of Economic Theory, 25(1), 92-100. https://doi.org/10.1016/0022-0531(81)90018-1The origin of the game.
- McKelvey, R. D., & Palfrey, T. R. (1992). An experimental study of the centipede game. Econometrica, 60(4), 803-836. https://doi.org/10.2307/2951567The first experiment: take rates rise toward the last move, and games end earlier with experience.
- Crosetto, P., & Mantovani, M. (2018). Representation effects in the centipede game. PLOS ONE, 13(10), e0204422. https://doi.org/10.1371/journal.pone.0204422A linear game kept on screen at every decision, with the current one marked.
- Krockow, E. M., Colman, A. M., & Pulford, B. D. (2016). Cooperation in repeated interactions: A systematic review of Centipede game experiments, 1992-2016. European Review of Social Psychology, 27(1), 231-282. https://doi.org/10.1080/10463283.2016.1249640What changes how people play: payoffs, repetition and social value orientation.
- Palacios-Huerta, I., & Volij, O. (2009). Field centipedes. American Economic Review, 99(4), 1619-1635. https://doi.org/10.1257/aer.99.4.161969% of highly ranked chess players stopped at once.
- Levitt, S. D., List, J. A., & Sadoff, S. E. (2011). Checkmate: Exploring backward induction among chess players. American Economic Review, 101(2), 975-990. https://doi.org/10.1257/aer.101.2.975Chess players almost never played the equilibrium; the game is ill-suited to testing backward induction.
- Ghosh, S., Heifetz, A., Verbrugge, R., & de Weerd, H. (2017). What drives people's choices in turn-taking games, if not game-theoretic rationality? Electronic Proceedings in Theoretical Computer Science, 251, 265-284. https://doi.org/10.4204/EPTCS.251.19Turn-taking games against a computer that participants know about.
Source papers
- McKelvey and Palfrey (1992): 58 students played 281 six-move games for money against one another, with piles that doubled at each pass. The chance of taking at each move was .01, .06, .21, .53, .73 and .85. Other points, partners and game lengths will give other rates.Paradigm source, Adults
Questions
How long does the Take or Pass Game take?
About 4 to 7 minutes, as listed in the test library.
What device does a participant need?
Untimed test: runs on desktops, laptops, tablets and phones. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Passes per Game, Take Rate, First Turn, Take Rate, Second Turn, Take Rate, Third Turn, Passes per Game, First Half, Passes per Game, Second Half, Change in Passes per Game, Points Won, Check Correct, Median Choice Time, Choices Under 100 ms and Games. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Take or Pass Game used in research?
34 papers from 2021 to 2025 name it in their title or abstract, down 21% compared with 2016 to 2020. That ranks it 343 of 495 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.22.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Take or Pass Game have?
In the published research, 0 papers report its test-retest reliability and 0 registered clinical studies use it as an outcome (evidence grade D in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Games (6 to 24), partner wording (A computer program (standard), Neutral: your partner) and payment line (Hypothetical points (standard), The study pays for points). Settings can be changed on yearly plans; free and pay-as-you-go studies use the standard settings.
Which languages can participants use?
Participants can take this test in English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese. English is on every plan, and the other languages come with yearly plans. Have a native speaker read the machine-translated text before launch.
Is this a clinical test?
No. ConductCognition is for research use only. It runs, scores and exports the test; clinical interpretation is outside the platform.
Every test: the full test list.
Run this test in your next study
Start free with 10 free participants on any test, then use participant credits or a yearly plan.