Decision Making
Two Gardens Game
Also called: two-step game, two-stage decision game, model-based learning game, gamified two-step task
Measure how far choices follow where each gate usually leads, in people aged 8 and over: two gates, two gardens with two bushes each, and berries that come and go, scored over two sessions.
8-15 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Preview any test free after you sign up.
- Papers, 2021 to 2025
- 70
- Evidence grade
- C
- Registered clinical studies
- 4
Figures for the Two-Step Task, the published test this game is compared with.
How these are countedWhat it measures
Measures how people aged 8 and over use where a gate leads to find berries, over two sessions, with an adult reading the screens to children of 8 and 9.
Constructs
- Model-based planning
- Learning from rewards
- Using where choices lead
- Executive function
Research uses
- Research on how people use what they know about where their choices lead, in adults and in children from 8
How the paradigm works
The Two Gardens Game measures how people aged 8 and over use where each of two gates usually leads while they learn from berries. It is a game in the family of the two-step tasks, in which a first choice leads to one of two second-stage states, usually the one it belongs to and now and then the other, and a second choice may give a reward whose chance changes slowly. Each trial shows two gates. A gate usually opens onto its own garden (7 trips in 10) and now and then onto the other. The garden shows two bushes, and a bush may give a berry; a berry buys nothing, and no screen shows a count. A part has 100 trials in two blocks of 50, with a rest between, and a check before it asks which garden each gate usually opens onto. The paper counts and grades on this page are those of the Two-Step Task, the published test the game is compared with, not the game's own.
More about the paradigm
The main score, the planning index, is the effect of a berry on staying with the same gate after a common trip minus its effect after a rare trip, over the pairs of consecutive trials of a set of two parts. It runs from minus 2 to 2, and choices that follow where the gates usually lead score higher. The reward index, the effect of a berry on staying, and the four stay probabilities are scored beside it. One session alone gives its part's own indices, and the planning index of one part is the figure the study view's split-half and retest read. The set's score needs both parts usable, one form and one look, and at least 10 pairs in each of the four cells; an interrupted trial leaves both pairs it is in. A part with the same side or the same gate on 95% or more of its choices, or with fewer than 80 usable trials, is marked unusable and gives no set score.
The game differs from the Two-Step Task in four ways: no time limit on a choice, a score pooled over two sessions, a check of the gardens before each part, and the same gardens in every session. The validation plan also compares it with Tower Planning. The two-step design follows Daw and colleagues (2011), and its 70% common trips, blocks of 50 with 15 rare trips and slowly drifting chances follow the Two-Step Task in this library. The check of the gardens follows the finding that many people misunderstand the task unless it is explained (Feher da Silva and Hare 2020). The gates, gardens, bushes and berries, the 8 forms, the arrangement of the drawings, the display times, the 100-trial parts and the quality cuts are our own.
The game is built for repeated visits and needs a study with visits, or two steps of the game in one sitting. Give the first part's visit a one-day window with no late days and the second part's visit a window that opens the day after it, for at most 7 days, with no late days; every window counts from the study's first visit. Two steps of the game in one sitting make a set 0 days apart. The session works out which part it is from the same participant's earlier sessions of the game: a second part follows a completed first part by at most 14 days, an abandoned or skipped part starts a new set on the next of 8 forms, and a reload keeps the part and its form. Add the game only at these visits, and the next set at the next time point. For a retest, run a set 7 to 28 days after the baseline set, and run two sets at each time point for a study that reads change within a person. A researcher's preview of one step shows a first part, and a protocol preview with two steps of the game shows a second part on the second step, scored without its set.
Scores
- Planning Indexscore
- Over the set's pairs of consecutive trials: the effect of a berry on staying with the same gate after a common trip minus its effect after a rare trip, from minus 2 to 2. Empty for a first part, and unless the set is complete, both parts are usable and each of the four cells holds at least 10 pairs.
- Higher means choices followed where the gates usually lead more; the score needs both parts of a set
- Reward Indexscore
- Over the set: the share of stays with the same gate after a berry on a common trip plus the share after a berry on a rare trip, minus the same two shares after no berry, from minus 2 to 2. Empty unless the set is complete.
- Higher means a berry made the person more likely to stay with the same gate; a value near 0 means berries did not guide choices
- Stay after a Berry on a Common Trip%
- Over the set: the share of trials after a berry on a common trip that were followed by the same gate.
- Informational
- Stay after a Berry on a Rare Trip%
- Over the set: the share of trials after a berry on a rare trip that were followed by the same gate.
- Informational
- Stay after No Berry on a Common Trip%
- Over the set: the share of trials after no berry on a common trip that were followed by the same gate.
- Informational
- Stay after No Berry on a Rare Trip%
- Over the set: the share of trials after no berry on a rare trip that were followed by the same gate.
- Informational
- Planning Index of One Partscore
- The planning index from this part's 100 trials alone, each of the four cells holding at least 5 pairs. A session gives it whether or not its set is complete.
- Higher means choices followed where the gates usually lead more; read within one study
- Reward Index of One Partscore
- The reward index from this part's trials alone.
- Informational; higher means berries guided choices more
- Set Completecount
- 1 when this session is the second part of a set whose first part could be read and the two parts agree on form, look and part, and both are usable; otherwise 0.
- Informational; the planning index is empty when this is 0
- Part of the Setcount
- 1 for a first part and 2 for a second part, as the session's own rows say.
- Informational
- Days since the First Partcount
- Whole calendar days from the first part's finish to this session's completion, in the participant's time zone. Empty for a first part.
- Informational
- Pairs in the Setcount
- The pairs of consecutive usable trials: both parts together when the set is complete, this part's alone otherwise.
- Informational; more pairs give a steadier index
- Answered Differently in the Two Partscount
- 1 for a complete set whose parts were answered differently: by touch in one part and by key or mouse in the other.
- Informational; a change of how the game was answered can move scores
- Berry Rate%
- The share of this part's usable trials in which the bush chosen gave a berry.
- Informational
- Share of Choices on the Better Bush%
- The share of this part's usable trials in which the bush chosen was the one with the higher chance of a berry in its garden.
- Informational; higher means the person found the bush that gave more berries
- Slowing after a Rare Tripms
- The median time to choose a bush after a rare trip minus the median after a common trip, in this part's usable trials.
- Informational; no score depends on a time
- Rounds of the Gardens Checkcount
- The rounds of the check before each part until both gates' gardens were named right, from 1 to 3, and 4 when they never were.
- Informational; the score does not depend on it, and a check never passed is flagged for review
- Usable Trialscount
- Test trials counted in the scores. Left out: a trial in which the tab was hidden, the window changed size, or a second finger or a palm touched the screen, and any row that does not match the form's plan.
- Informational; fewer than 80 of the 100 make the part unusable
- Share of Key Choices%
- The share of the scored gate choices made with a key (Enter or Space on a focused gate) and not a tap or a click.
- Informational; the comparison of touch and keys reads this share within each look
- Share of Touch Choices%
- The share of the scored gate choices made by touch.
- Informational; the comparison of touch and keys reads this share within each look
Use and evidence
70 papers from 2021 to 2025 name this task in their title or abstract, up 37% compared with 2016 to 2020. That ranks it 373 of 711 test families in our census. These counts belong to the Two-Step Task, the published test this game is compared with. This game is a new measure, so its own reliability and norms come from use. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 23 |
| 2024 | 22 |
| 2023 | 8 |
| 2022 | 7 |
| 2021 | 10 |
| 2020 | 17 |
| 2019 | 11 |
| 2018 | 9 |
| 2017 | 9 |
| 2016 | 5 |
| 2015 | 5 |
| 2014 | 2 |
| 2013 | 2 |
| 2012 | 3 |
| 2011 | 1 |
| 2010 | 1 |
| 2009 | 0 |
| 2008 | 0 |
| 2007 | 0 |
| 2006 | 0 |
| 2005 | 0 |
| 2004 | 0 |
| 2003 | 1 |
| 2002 | 0 |
| 2001 | 0 |
| 2000 | 0 |
Use grade D
30 to 99 papers from 2021 to 2025.
Evidence grade C
At least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies.
- Papers on test-retest reliability
- 0
- Papers on reliability or validity
- 17
- Registered clinical studies using it as an outcome
- 4
Measured on ConductCognition
Pooled across the studies that feed our platform norms, by device group. A figure needs 50 or more people.
No pooled figure yet.
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 the Two-Step Task, the published test this game is compared with, not the game's own.
Source: ConductCognition test census 0.58.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Gates, gardens and bushes
- Two gates stand side by side. A gate usually opens onto its own garden (7 trips in 10) and now and then onto the other. Each garden has two bushes. Everything is drawn in code, with no picture file: the gates differ in shape (arched and square), the gardens in ground (stripes and dots) and the four bushes in outline (round, tall, wide and pointed), in colors that stay apart under simulated color blindness. Each person sees one of 8 arrangements of these drawings, drawn once for them and the same at every session.
- A trial
- The player picks a gate, which glows for 400 ms and opens over 600 ms onto a garden. The garden's two bushes appear side by side and the player picks one, which shows a berry or no berry for 1,000 ms. The next trial starts 500 ms later. There is no time limit on either choice. Each gate is on the left on 25 of the 50 trials of a block, and each bush on the left on half of its garden's trials.
- Berries
- Each bush's chance of a berry drifts slowly between 0.25 and 0.75, in steps with a standard deviation of 0.025. The chances come from our own seeded random walks, and the berries are drawn from them once, in advance, and stored with the form, so every player meets the same berries on the same trials. A berry buys nothing: there is no points count and no tally.
- Parts and sets
- A part has 100 trials in 2 blocks of 50, with a rest between. Each block holds 15 trips to the other garden and starts with a trip to the gate's own garden. A part takes about 8 to 10 minutes for an adult and up to 15 for a child of 8 or 9. A set is two parts, played as two sessions of one study up to 14 days apart. The main score takes both parts, and one part alone gives its own indices.
- Check and practice
- Before each part, a check asks for each gate which garden it usually opens onto. A wrong answer shows the right garden, and the check runs again, up to 3 rounds, and then the game goes on. Then come 10 practice trials with their outcomes, in which every bush has a chance of 0.5 and 3 trips go to the other garden. The check and the practice are the same in both parts.
- Forms
- 8 forms with the same trips and sides, in which each bush takes a different one of the four chance series: the two bushes of one garden swapped, the two bushes of the other swapped, or the two gardens' pairs swapped. Each person gets the forms in an order drawn once for them, and both parts of a set use the same form.
- Gates and bushes on screen
- Every gate, garden and bush is at least 120 px and is laid out from the window before each trial: 156 px on a 360 by 640 phone and 280 px on a tablet or a computer. Screen buttons are 64 px tall. The first tap on a gate or bush chooses, and a second tap on it within 300 ms changes nothing and is counted. Enter or Space on a focused gate or bush chooses too, unless the key was held down from before the choice opened. Older-adult mode makes the words 1.2 times larger and the buttons 77 px tall, and changes no picture, timing or score. The game look adds a sky and a path at the edges of the screen, and the plain look draws the same pictures on white.
- Age and reader
- For people aged 8 and over. An adult sits beside a child of 8 or 9, reads each screen aloud and does not suggest a gate or a bush. For children of 8 to 12 the score is read for groups by age, not for one child.
- Schedule
- The game needs a study with visits, or two steps of the game in one sitting. Give the first part's visit a one-day window with no late days, and the second part's visit a window that opens the day after it, for at most 7 days, with no late days; every window counts from the study's first visit. Two steps in one sitting make a set 0 days apart. A second part follows a completed first part by at most 14 days; after that, or after an abandoned or skipped part, a new set begins on the next of 8 forms. Add the game only at these visits, and the next set at the next time point. For a retest, run a set 7 to 28 days after the baseline set; a study that reads change within a person needs two sets at each time point. Keep each person on one device class and one look at every visit, and give each person their own link, because a link shared by two people mixes their parts.
- Scores
- The planning index (the main score, from minus 2 to 2): the effect of a berry on staying with the same gate after a common trip minus its effect after a rare trip, over the pairs of consecutive trials of both parts. It needs both parts usable, one form and one look, and at least 10 pairs in each of the four cells. Also given: the reward index (the effect of a berry on staying), the four stay probabilities, each part's own planning and reward indices (at least 5 pairs in each cell), the berry rate, the share of choices on the bush with the higher chance, the median time to choose a bush after a rare trip minus after a common one, the rounds of the check, the days since the first part and the number of usable trials. A trial is left out, with both pairs it is in, when the tab was hidden, the window changed size, or a second finger or a palm touched the screen. A part with fewer than 80 usable trials, or with the same side or the same gate on 95% or more of its choices, is marked unusable and gives no set score.
The design follows the two-step task of Daw and colleagues (2011), as the Two-Step Task in this library does: 70% common trips, blocks of 50 with 15 rare trips and chances that drift slowly. It adds a check of the gardens before each part, because many people misunderstand the task unless it is explained (Feher da Silva and Hare 2020), and it scores a set of two parts of 100 trials, where a smartphone two-step game scored sessions of 200 trials (Donegan and colleagues 2023). The gates, gardens, bushes, berries, forms, check and looks are our own. This game is built for repeated visits and is a new measure for people aged 8 and over: its reliability and norms come from use, the figures published in those studies are not its own, and it computes no percentiles. It needs a study with visits, or two steps of the game in one sitting.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: two_gardens_game test trials of one part, a gate choice and a bush choice each, held to the form's plan, leaving out a trial made with the tab hidden, the window resized, or a second finger or a palm on the screen. The same side or the same gate on 95% or more of a part's gate choices, or fewer than 80 scored trials of the 100, invalidates the part; 25 or more gate choices faster than 200 ms, a gardens check never passed, a visit not given, a second part without a usable first part, or a set answered in two different ways, flags it for review.
Why these rules
No gate and no bush is right in advance: what the game measures is whether the next gate choice follows the last berry and trip, so a person who always takes one gate shows a habit and not the paradigm, and the smartphone two-step game this one follows (Donegan and colleagues 2023) dropped players who chose the same option on over 95% of trials. Choosing by side alone means the gates were not looked at (the library Two-Step Task's rule, 95%). A part with fewer than 80 of its 100 trials scored is too thin for stay pairs, as that game dropped players who missed over 20% of their first session's trials. An answer under 200 ms is faster than a simple visual reaction (the platform's untimed-choice rule). Many people misconceive the task unless it is explained, so a gardens check that is never passed is flagged for review. The planning index rests on a set of two parts with the same form and look, and a change of how the game was answered between them is flagged because a device change moves scores (ours). A run that holds two forms, parts or looks cannot be read against one plan and is not scored.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Look | Game | Game, Plain |
Every plan can change the look, which never changes how a result is scored.
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 Two Gardens Game in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Daw, Gershman, Seymour, Dayan and Dolan, 2011), with the standard settings (look: Game). The main score was planning index (over the set's pairs of consecutive trials: the effect of a berry on staying with the same gate after a common trip minus its effect after a rare trip, from minus 2 to 2).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Daw, N. D., Gershman, S. J., Seymour, B., Dayan, P., & Dolan, R. J. (2011). Model-based influences on humans' choices and striatal prediction errors. Neuron, 69(6), 1204-1215. https://doi.org/10.1016/j.neuron.2011.02.027The task this game adapts: a first choice leads to one of two second-stage states, usually the one it belongs to, and a second choice may give a reward whose chance drifts slowly; staying with the first choice is analyzed by reward and by whether the trip was common or rare (abstract read).
- Feher da Silva, C., & Hare, T. A. (2020). Humans primarily use model-based inference in the two-stage task. Nature Human Behaviour, 4(10), 1053-1066. https://doi.org/10.1038/s41562-020-0905-yChoices were mainly model-based when the instructions were more detailed, and many people misunderstand the task without them, so each part opens with a check of the gardens (abstract read).
- Donegan, K. R., Brown, V. M., Price, R. B., Gallagher, E., Pringle, A., Hanlon, A. K., & Gillan, C. M. (2023). Using smartphones to optimise and scale-up the assessment of model-based planning. Communications Psychology, 1, 31. https://doi.org/10.1038/s44271-023-00031-yReliability of the model-based index in a smartphone two-step game, with 100 trials in a block and 200 in a session (abstract read; the trial counts are from the full text).
- Decker, J. H., Otto, A. R., Daw, N. D., & Hartley, C. A. (2016). From creatures of habit to goal-directed learners: Tracking the developmental emergence of model-based reinforcement learning. Psychological Science, 27(6), 848-858. https://doi.org/10.1177/0956797616639301A two-step task given to children, adolescents and adults, the source for including children from 8 (abstract read).
Source papers
Questions
How long does the Two Gardens Game take?
About 8 to 15 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?
Planning Index, Reward Index, Stay after a Berry on a Common Trip, Stay after a Berry on a Rare Trip, Stay after No Berry on a Common Trip, Stay after No Berry on a Rare Trip, Planning Index of One Part, Reward Index of One Part, Set Complete, Part of the Set, Days since the First Part, Pairs in the Set, Answered Differently in the Two Parts, Berry Rate, Share of Choices on the Better Bush, Slowing after a Rare Trip, Rounds of the Gardens Check, Usable Trials, Share of Key Choices and Share of Touch Choices. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Two-Step Task used in research?
These counts belong to the Two-Step Task, the published test this game is compared with. This game is a new measure, so its own reliability and norms come from use. 70 papers from 2021 to 2025 name the task in their title or abstract, up 37% compared with 2016 to 2020. That ranks it 373 of 711 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.58.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Two-Step Task have?
These counts belong to the Two-Step Task, the published test this game is compared with. This game is a new measure, so its own reliability and norms come from use. In the published research, 0 papers report the task's test-retest reliability and 4 registered clinical studies use it as an outcome (evidence grade C in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Look (Game, Plain). Every plan can change the look, which never changes how a result is scored.
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.
Related tests