Decision Making
Helicopter Task
Also called: predictive inference task, change-point task, helicopter bag-drop task, adaptive learning task
Measure adaptive learning in a changing world: participants move a bucket to catch bags dropped by a hidden helicopter that sometimes moves, scored as how much the learning rate rises after likely changes.
8-12 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
- 22
- Evidence grade
- C
- Registered clinical studies
- 2
What it measures
Measures how people adjust learning in a changing world. A hidden helicopter drops bags that scatter around it and now and then moves to a new place; before each drop the participant moves a bucket to where the next bag will fall. The main score is how much the learning rate rises when a miss signals that the helicopter has probably moved (the change-point weight of the normative model).
Constructs
- Adaptive learning
- Belief updating
- Change-point detection
- Learning under uncertainty
Research uses
- Computational psychiatry research on learning rates
- Aging and development of adaptive learning
- Within-study comparison of belief updating after surprises
How the paradigm works
Places run from 0 to 300 units, as in the sources. The wind spreads the bags with a standard deviation of 10 units in calm blocks and 25 in windy ones, and the bucket is 3 spreads wide, as in McGuire and colleagues' task. Each block holds 40 drops in five runs of 4, 6, 8, 10 and 12 drops in a seeded order, so the helicopter moves four times in every block, each time to a place drawn at random from the whole field. The standard 4 blocks alternate calm and windy. The bucket moves by a click, a tap, a drag or the arrow keys, and Drop has no time limit. Practice comes first: 10 drops with the helicopter in view. Each bag caught earns 1 point (0 to 10 in the settings), and the study decides what points are worth. About 8 to 12 minutes.
Scores
- Change-Point Weightscore
- Weight on the miss times the change-point probability in the regression of each move of the bucket (McGuire and colleagues, 2014).
- Near 1 when learning rises with the chance that the helicopter has moved, as in the normative model; near 0 when it does not
- Uncertainty Weightscore
- Weight on the miss times relative uncertainty, outside likely moves, in the same regression.
- Near 1 when learning follows uncertainty about the helicopter's place, as in the normative model
- Fixed Learning Ratescore
- Weight on the miss alone in the same regression: the share of each miss the bucket moves whatever the model's view.
- Near 0 in the normative model
- Mean Learning Ratescore
- Slope of the moves of the bucket on the misses, over all scored drops.
- Informational; 1 means moving all the way to each bag
- Learning Rate After a Movescore
- Median of move over miss after the first bag from a new place of the helicopter, for misses of at least 5 units.
- Near 1 when a move of the helicopter is followed at once
- Learning Rate Between Movesscore
- Median of move over miss after the other bags, for misses of at least 5 units.
- Lower means more of the wind is ignored; above 1 is flagged
- Mean Learning Rate, Calmscore
- Slope of the moves on the misses in the calm blocks.
- Informational
- Mean Learning Rate, Windyscore
- Slope of the moves on the misses in the windy blocks.
- Informational; usually lower than in calm blocks
- Estimation Errorscore
- Mean distance from the bucket to the hidden helicopter, in units of the 300-unit field, leaving out each block's first drop.
- Lower is closer to where the helicopter was
- Prediction Errorscore
- Mean distance from the bucket to the bag, in units of the 300-unit field, leaving out each block's first drop.
- Lower is better; the wind sets a floor
- Perseveration%
- Share of drops after a block's first on which the bucket was not moved at all.
- Informational; 90% or more is flagged
- Model Fitscore
- R-squared of the regression that gives the three weights.
- Below 0.2 is flagged: the moves barely follow the misses
- Bags Caught%
- Share of test bags that landed in the bucket.
- Informational
- Points Wonpoints
- Points for bags caught.
- Informational
- Median Drop Timems
- Median time from a round appearing to Drop.
- Informational; the task has no time limit
- Moves Outside the Spancount
- Moves that ended more than 15 units outside the span from the old bucket to the bag; left out of the learning measures.
- Quality check
- Anticipationscount
- Presses of Drop within 100 ms of a round appearing. They are not taken.
- Quality check
- Drops Madecount
- Test drops in the scored run.
- Fewer than planned is flagged for review
- Planned Dropscount
- Test drops the settings ask for: 40 per block.
- Informational
Use and evidence
22 papers from 2021 to 2025 name this task in their title or abstract. That ranks it 325 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 7 |
| 2024 | 8 |
| 2023 | 2 |
| 2022 | 2 |
| 2021 | 3 |
| 2020 | 4 |
| 2019 | 2 |
| 2018 | 0 |
| 2017 | 2 |
| 2016 | 0 |
| 2015 | 0 |
| 2014 | 0 |
| 2013 | 0 |
| 2012 | 3 |
| 2011 | 3 |
| 2010 | 0 |
| 2009 | 0 |
| 2008 | 0 |
| 2007 | 0 |
| 2006 | 0 |
| 2005 | 0 |
| 2004 | 0 |
| 2003 | 1 |
| 2002 | 1 |
| 2001 | 1 |
| 2000 | 1 |
Use grade E
Fewer than 30 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
- 4
- Papers on reliability or validity
- 7
- Registered clinical studies using it as an outcome
- 2
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 0.14.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Field
- Places from 0 to 300 units along the ground. The bucket is 3 spreads of the wind wide.
- Bags
- Each bag lands at the helicopter's place plus wind: a normal spread of 10 units in calm blocks and 25 in windy ones, rounded and kept on the field.
- Moves
- Each block holds 40 drops in runs of 4, 6, 8, 10 and 12 drops in a seeded order. After each run the helicopter moves to a new place drawn at random from the whole field.
- Blocks
- 4 blocks (standard; 2 to 4), alternating calm and windy, the first drawn from the session's seed, with a rest between blocks. The bucket starts each block in the middle.
- Drop
- The bucket moves by a click, a tap, a drag or the arrow keys, and Drop has no time limit. The bag and the red line of the miss stay in view for 1,000 ms. Presses of Drop under 100 ms are not taken.
- Points
- 1 point (standard; 0 to 10) for each bag caught. The points are hypothetical by default; a setting changes the instruction line to say the study pays.
- Practice
- 10 drops in the calm wind with the helicopter in view. It moves once, after 5 drops.
- Model
- Change-point probability and relative uncertainty come from the reduced Bayesian model with a hazard rate of 0.1, run on each block's bags.
The sources used a joystick or keys, hundreds of drops and long training. This test uses its own schedule of moves, a bucket moved by pointer or keys and a short practice, so its weights are not directly comparable with theirs. The scores support research 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: predictive_inference_test from the scored run, one per drop. A bucket's place has no correct option to compare with chance, so accuracy is not judged against chance. Drops within 200 ms of the round appearing on at least half the drops invalidate. Not moving the bucket on 90% of drops or more, a model fit below R-squared 0.2, moves far outside the span from the last bucket to the last bag on 25% of drops or more, a median learning rate above 1 between moves of the helicopter, and a run short of its planned drops are flagged for review and withhold the percentile.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. 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. The learning measures come from how the bucket moves after each bag, so a bucket that never moves, moves that do not follow the misses, moves that land far from both the last bucket and the last bag, or moves that keep overshooting the bag leave them with little to measure. McGuire and colleagues (2014) left out a participant whose median learning rate was above 1, and Nassar and colleagues (2016) left out the 1.1% of moves that fell far from any delta-rule update. Each is flagged rather than excluded, because some groups do keep the bucket still (Bruckner and colleagues, 2025).
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Blocks | 4 | 2 to 4 |
| Points per bag caught | 1 | 0 to 10 |
| 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 and tablets. 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 Helicopter Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Nassar, Wilson, Heasly & Gold, 2010; McGuire, Nassar, Gold & Kable, 2014), with the standard settings (blocks: 4; points per bag caught: 1; payment line: Hypothetical points). The main score was change-point weight (weight on the miss times the change-point probability in the regression of each move of the bucket (McGuire and colleagues, 2014)).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Nassar, M. R., Wilson, R. C., Heasly, B., & Gold, J. I. (2010). An approximately Bayesian delta-rule model explains the dynamics of belief updating in a changing environment. Journal of Neuroscience, 30(37), 12366-12378. https://doi.org/10.1523/JNEUROSCI.0822-10.2010The task's model: change-point probability and relative uncertainty set the learning rate.
- McGuire, J. T., Nassar, M. R., Gold, J. I., & Kable, J. W. (2014). Functionally dissociable influences on learning rate in a dynamic environment. Neuron, 84(4), 870-881. https://doi.org/10.1016/j.neuron.2014.10.013The helicopter version, and the regression of each move on the model's terms.
- Nassar, M. R., Bruckner, R., Gold, J. I., Li, S.-C., Heekeren, H. R., & Eppinger, B. (2016). Age differences in learning emerge from an insufficient representation of uncertainty in older adults. Nature Communications, 7, 11609. https://doi.org/10.1038/ncomms11609Older adults made less use of uncertainty when learning.
- Bruckner, R., Nassar, M. R., Li, S.-C., & Eppinger, B. (2025). Differences in learning across the lifespan emerge via resource-rational computations. Psychological Review, 132(3), 556-580. https://doi.org/10.1037/rev0000526Perseveration and estimation error from childhood to older age.
- Loosen, A. M., Seow, T. X. F., & Hauser, T. U. (2024). Consistency within change: Evaluating the psychometric properties of a widely used predictive-inference task. Behavior Research Methods, 56(7), 7410-7426. https://doi.org/10.3758/s13428-024-02427-yInternal consistency and test-retest reliability of the task's measures online.
Source papers
- McGuire JT, Nassar MR, Gold JI, Kable JW (2014). Functionally dissociable influences on learning rate in a dynamic environment. Neuron, 84(4):870-881.N 32, 18 to 30
- No normative data for this task.Paradigm source, N/A
Questions
How long does the Helicopter Task take?
About 8 to 12 minutes, as listed in the test library.
What device does a participant need?
Untimed test: runs on desktops, laptops and tablets. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Change-Point Weight, Uncertainty Weight, Fixed Learning Rate, Mean Learning Rate, Learning Rate After a Move, Learning Rate Between Moves, Mean Learning Rate, Calm, Mean Learning Rate, Windy, Estimation Error, Prediction Error, Perseveration, Model Fit, Bags Caught, Points Won, Median Drop Time, Moves Outside the Span, Anticipations, Drops Made and Planned Drops. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Helicopter Task used in research?
22 papers from 2021 to 2025 name it in their title or abstract. That ranks it 325 of 395 test families in the ConductCognition test census (use grade E). Source: ConductCognition test census 0.14.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Helicopter Task have?
In the published research, 4 papers report its test-retest reliability and 2 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?
Blocks (2 to 4), points per bag caught (0 to 10) 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.