Decision Making
Weather Prediction Task
Also called: WPT, probabilistic classification learning, probabilistic category learning task, weather forecasting task
Measure probabilistic learning: participants forecast sun or rain from cards that each predict the weather only part of the time, scored as the share of optimal forecasts.
8-11 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
- 25
- Evidence grade
- C
- Registered clinical studies
- 5
What it measures
Measures probabilistic learning from feedback. On each trial one to three of four cards appear and participants forecast sun or rain, then see the weather. Each card predicts the weather only part of the time, so learning builds over trials. The main score is the share of forecasts that chose the more likely weather for the cards shown, over 3 blocks of 50.
Constructs
- Probabilistic category learning
- Learning from feedback
- Cue weighting
- Strategy use
Research uses
- Research on striatal and declarative learning
- Parkinson's disease, schizophrenia and developmental research
- Within-study comparison of learning curves
How the paradigm works
The weather prediction task measures probabilistic classification learning: learning from feedback which outcome a set of cues predicts when no cue predicts it for sure (Knowlton, Squire and Gluck, 1994). Patients with Parkinson's or Huntington's disease learned it poorly from the start of training, while amnesic patients learned normally over the first 50 trials, which linked early learning to the striatum, as Gluck, Shohamy and Myers (2002) summarize. Later work found that people solve the task with different strategies, many of them simple, so the task does not measure implicit learning alone (Gluck, Shohamy and Myers, 2002; Bochud-Fragnière, Banta Lavenex and Lavenex, 2022).
More about the paradigm
Each trial shows one, two or three of four cards, each in its own place. The participant forecasts sun or rain by clicking or tapping, or with the F and J keys, with no time limit; after 5 seconds a line asks for an answer. The weather then shows for 1.5 seconds with Correct or Wrong, and the cards stay on screen. Sun and rain are equally likely, and each card appears independently of the others with a chance that depends on the weather: when it is sunny the four cards appear on 80%, 60%, 40% and 20% of trials, and when it rains the reverse. Trials with no card or all four are left out, so each card predicts sun on roughly 80%, 60%, 40% or 20% of the trials on which it appears. A session has 3 blocks of 50 trials (2 to 4 in the settings) with a rest between them, after 4 practice trials with two practice cards. Which card carries which chance, and the order of the trials, come from the session's seed. The cards were drawn by code for this task, and the trial counts are computed by code, not taken from a published table. Each right forecast earns 1 point (0 to 10 in the settings; 0 runs without points), and the study decides what points are worth.
The main score is the optimal choice rate: the share of forecasts that chose the more likely weather for the cards shown, whatever the weather turned out to be, as Gluck and colleagues (2002) score the task. The two card combinations that predict sun and rain equally are left out. The task also reports the rate in each block, the gain from the first block to the last, the share of forecasts that matched the weather, and the strategy scores of Gluck and colleagues for the last block: answering by all cards, by one card, or by the single-card combinations alone. Scores support research and do not diagnose.
Scores
- Optimal Choice Rate%
- Share of forecasts that chose the more likely weather for the cards shown, whatever the weather turned out to be. Combinations that predict sun and rain equally are left out.
- Higher is better; 0.5 is chance
- Optimal Choices, Block 1%
- The optimal choice rate in block 1.
- Higher is better
- Optimal Choices, Block 2%
- The optimal choice rate in block 2.
- Higher is better
- Optimal Choices, Block 3%
- The optimal choice rate in block 3. Blank when the session had 2 blocks.
- Higher is better
- Optimal Choices, Block 4%
- The optimal choice rate in block 4. Blank when the session had fewer blocks.
- Higher is better
- Learning Gain%
- The optimal choice rate in the last block minus that in block 1.
- Positive values mean forecasts improved during the session
- Forecasts Matching the Weather%
- Share of forecasts that matched the weather shown. Even the best forecaster is wrong on some trials.
- Higher is better
- Multi-Cue Strategy Scorescore
- How far the last block's forecasts are from always choosing the more likely weather for the cards shown (Gluck, Shohamy and Myers, 2002), from 0 for a perfect fit to 1. Blank with fewer than 20 forecasts in that block.
- Lower means a closer fit; under 0.1 was the source's criterion
- One-Cue Strategy Scorescore
- The same score for answering by one card alone, the best of the four cards.
- Lower means a closer fit
- Singleton Strategy Scorescore
- The same score for answering right on the single-card combinations and guessing on the others.
- Lower means a closer fit
- Multi-Cue Strategy Fits Best0/1
- 1 when the multi-cue strategy has the lowest of the three scores.
- The share of 1s is the share of participants best fit by the multi-cue strategy
- Points Wonpoints
- Points for forecasts that matched the weather.
- Informational
- Median Answer Timems
- Median time from the cards' onset to the forecast.
- Informational; the task has no time limit
- Forecasts After the Remindercount
- Forecasts made after the 5 second reminder.
- Informational
- Anticipationscount
- Presses within 100 ms of the cards' onset. They are not taken as forecasts.
- Quality check
- Forecasts Madecount
- Test forecasts in the scored run.
- Fewer than planned is flagged for review
- Planned Forecastscount
- Test trials in the session: 50 per block.
- Informational
Use and evidence
25 papers from 2021 to 2025 name this task in their title or abstract, up 4% compared with 2016 to 2020. That ranks it 317 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 7 |
| 2024 | 4 |
| 2023 | 7 |
| 2022 | 1 |
| 2021 | 6 |
| 2020 | 4 |
| 2019 | 4 |
| 2018 | 5 |
| 2017 | 5 |
| 2016 | 6 |
| 2015 | 12 |
| 2014 | 12 |
| 2013 | 10 |
| 2012 | 11 |
| 2011 | 12 |
| 2010 | 6 |
| 2009 | 6 |
| 2008 | 5 |
| 2007 | 6 |
| 2006 | 5 |
| 2005 | 8 |
| 2004 | 5 |
| 2003 | 7 |
| 2002 | 3 |
| 2001 | 0 |
| 2000 | 2 |
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
- 0
- Papers on reliability or validity
- 6
- Registered clinical studies using it as an outcome
- 5
Note on the counts: may include forecasting papers.
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 31 October 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Cards
- 4 cards drawn by code for this task, each with its own pattern and ink (stripes, dots, waves and checks) and each always in its own place. One, two or three show on each trial.
- Weather
- Sun and rain are equally likely. When it is sunny the four cues appear on 80%, 60%, 40% and 20% of trials, each independently of the others, and when it rains the reverse; trials with no card or all four are left out. In the standard session the cues come with sun on 81%, 61%, 39% and 19% of the trials on which they appear.
- Cards and cues
- Which card carries which cue is drawn from the session's seed and saved on every trial.
- Trials
- The expected count of each card combination with each weather, rounded so sun and rain stay equal, then dealt evenly into the blocks: 3 blocks (standard; 2 to 4) of 50 trials with a rest between them, each block in a seeded order.
- Trial
- The cards stay until the forecast, Sun or Rain, by click, tap, or the F and J keys, with no time limit; after 5,000 ms a line asks for an answer. Then Correct or Wrong and the weather for 1,500 ms, with the cards still shown, and a 500 ms blank.
- Optimal forecast
- The more likely weather for the cards shown, whatever the weather turns out to be. Where one weather is more likely, choosing it matches the weather on 88% of the standard session's trials.
- Points
- 1 point (standard; 0 to 10) for each forecast that matches the weather, with the total shown. 0 runs the task without points. The study decides what points are worth.
- Practice
- 4 trials with two practice cards, one that always comes with sun and one that always comes with rain.
Knowlton and colleagues used tarot cards, and Gluck and colleagues a fixed set of 200 trials. This test uses its own cards and computes its trial counts from the same kind of structure, so its rates are not directly comparable with theirs. People solve the task with different strategies, so the strategy scores help interpret the main score. 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: weather_prediction_test from the scored run, a forecast of sun or rain with no time limit. A forecast is right when it names the more likely weather for the cards shown (chance 50%); the combinations that predict both equally have no right answer. Accuracy below chance invalidates and accuracy not above chance is flagged. One answer on 90% of forecasts invalidates when accuracy is not above chance; sun and rain are equally frequent. Answers under 200 ms on at least half the forecasts invalidate. A run short of its planned forecasts is flagged for review and withholds the percentile.
Why these rules
Accuracy significantly below chance (one-sided exact test, p < .05) cannot come from guessing; it means the response mapping was reversed or the task was not followed (Slick, Sherman and Iverson 1999 treat below-chance forced-choice performance as definite response bias). Chance-level accuracy can be a real result in impaired participants, so it is only flagged. One option on at least 90% of answered trials, when the correct options are balanced and accuracy is not above chance, means the answers did not depend on the items (the task counterpart of long-string responding in surveys; Curran 2016). 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. Early forecasts are guesses by design, so only accuracy over the whole run is judged.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Blocks | 3 | 2 to 4 |
| Points per right forecast | 1 | 0 to 10 |
| Payment line | No mention of money (standard) | No mention of money (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.
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Weather Prediction Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Knowlton, Squire & Gluck, 1994; Gluck, Shohamy & Myers, 2002), with the standard settings (blocks: 3; points per right forecast: 1; payment line: No mention of money). The main score was optimal choice rate (share of forecasts that chose the more likely weather for the cards shown, whatever the weather turned out to be).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Knowlton, B. J., Squire, L. R., & Gluck, M. A. (1994). Probabilistic classification learning in amnesia. Learning & Memory, 1(2), 106-120. https://doi.org/10.1101/lm.1.2.106
- Gluck, M. A., Shohamy, D., & Myers, C. (2002). How do people solve the "weather prediction" task?: Individual variability in strategies for probabilistic category learning. Learning & Memory, 9(6), 408-418. https://doi.org/10.1101/lm.45202Cues of 0.8, 0.6, 0.4 and 0.2 associated with the weather independently, scoring by the optimal forecast, and the strategy scores.
- Knowlton, B. J., Mangels, J. A., & Squire, L. R. (1996). A neostriatal habit learning system in humans. Science, 273(5280), 1399-1402. https://doi.org/10.1126/science.273.5280.1399Impaired learning in Parkinson's disease from early in training.
- Bochud-Fragnière, E., Banta Lavenex, P., & Lavenex, P. (2022). What is the weather prediction task good for? A new analysis of learning strategies reveals how young adults solve the task. Frontiers in Psychology, 13, 886339. https://doi.org/10.3389/fpsyg.2022.886339Strategies people use to solve the task, in a 100-trial version.
Source papers
- Gluck MA, Shohamy D, Myers C (2002). How do people solve the "weather prediction" task? Individual variability in strategies for probabilistic category learning. Learning & Memory, 9(6):408-418.N 60, Young adults
- No normative data for this task.Paradigm source, N/A
Questions
How long does the Weather Prediction Task take?
About 8 to 11 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?
Optimal Choice Rate, Optimal Choices, Block 1, Optimal Choices, Block 2, Optimal Choices, Block 3, Optimal Choices, Block 4, Learning Gain, Forecasts Matching the Weather, Multi-Cue Strategy Score, One-Cue Strategy Score, Singleton Strategy Score, Multi-Cue Strategy Fits Best, Points Won, Median Answer Time, Forecasts After the Reminder, Anticipations, Forecasts Made and Planned Forecasts. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Weather Prediction Task used in research?
25 papers from 2021 to 2025 name it in their title or abstract, up 4% compared with 2016 to 2020. That ranks it 317 of 395 test families in the ConductCognition test census (use grade E). Source: ConductCognition test census 0.14.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Weather Prediction Task have?
In the published research, 0 papers report its test-retest reliability and 5 registered clinical studies use it as an outcome (evidence grade C in the ConductCognition test census).
Which settings can a researcher change?
Blocks (2 to 4), points per right forecast (0 to 10) and payment line (No mention of money (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.