Episodic Memory
Hypercorrection Task
Also called: hypercorrection effect task, high-confidence error correction paradigm, confidence and corrective feedback task
Measure whether high-confidence errors are more likely to be corrected after seeing the right answer.
8-12 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
- 3
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures whether errors made with higher confidence are more likely to be corrected after feedback. Participants answer fact questions, rate confidence, see each correct answer, and answer the questions again.
Constructs
- Metacognition
- Error correction
- Learning from feedback
Research uses
- Metacognitive monitoring research
- Memory updating research
- Within-study learning comparisons
How the paradigm works
An error made with strong confidence can be easier to correct than a tentative wrong answer. The Hypercorrection Task measures this relation by recording confidence before the correct answer appears, then testing the same knowledge again.
More about the paradigm
Participants answer 24 original fact questions, each with four numeric alternatives, rate confidence from 1 to 6 and see the correct answer for 3 seconds. After 30 seconds of arithmetic they answer the same questions in a new order with reshuffled alternatives. Two separate practice questions introduce the format. The question count, feedback duration and delay are bounded settings. Questions are written in all eight app languages.
The primary score is Goodman–Kruskal gamma between confidence in an initial error and whether that error is corrected at retest. Positive values mean greater correction of higher-confidence errors. Scores also include high- and low-confidence correction rates, their difference and their denominators. If all initial answers are correct, confidence does not vary, or all errors have the same retest outcome, gamma is undefined. Counts and accuracy remain available.
Scores
- Hypercorrection Gammascore
- Goodman-Kruskal gamma between initial confidence and retest correctness, among initial errors with both observations.
- Positive means higher-confidence errors were corrected more often; null when undefined
- High Minus Low Confidence Correction%
- High-confidence minus low-confidence error correction rate, in percentage points (difference multiplied by 100).
- Positive indicates more correction of high-confidence errors
- High-Confidence Error Correction%
- Correction rate among errors rated 4, 5 or 6.
- Read with the high-confidence error count
- Low-Confidence Error Correction%
- Correction rate among errors rated 1, 2 or 3.
- Read with the low-confidence error count
- Error Correction Rate%
- Share of paired initial errors with a confidence rating answered correctly at retest.
- Higher means more corrections in this task
- Initial Accuracy%
- Share of valid first-round answers correct.
- Initial knowledge context
- Retest Accuracy%
- Share of valid retest answers correct.
- Knowledge after feedback
- Mean Confidencepoints
- Mean initial confidence on the 1 to 6 scale.
- Read alongside accuracy
- Initial Errorscount
- Number of valid first answers that were wrong.
- Potential observations for hypercorrection
- Paired Errorscount
- Initial errors with valid confidence and retest answers.
- Denominator of the association
- High-Confidence Errorscount
- Paired errors rated 4 to 6.
- Denominator of high-confidence correction
- Low-Confidence Errorscount
- Paired errors rated 1 to 3.
- Denominator of low-confidence correction
- Corrected Errorscount
- Paired errors answered correctly at retest.
- Number of corrections
- Paired Questionscount
- Questions with valid first and retest answers.
- Completion context
- First Answerscount
- Valid initial answer choices.
- Completion context
- Retest Answerscount
- Valid retest answer choices.
- Completion context
- Confidence Ratingscount
- Valid initial ratings from 1 to 6.
- Completion context
- Planned Questionscount
- Number of selected questions.
- Setting context
- Median Answer Timems
- Median answer time across both rounds, excluding responses under 100 ms.
- Descriptive timing only
- Answers Under 100 mscount
- Answers given within 100 ms across both rounds.
- Retained as choices; speed is checked separately
- Earlier Sessionscount
- Previous sessions of this task, when scoring counted them.
- Prior exposure is flagged
Use and evidence
3 papers from 2021 to 2025 name this task in their title or abstract. That ranks it 481 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 1 |
| 2024 | 1 |
| 2023 | 0 |
| 2022 | 1 |
| 2021 | 0 |
| 2020 | 0 |
| 2019 | 2 |
| 2018 | 2 |
| 2017 | 1 |
| 2016 | 1 |
| 2015 | 5 |
| 2014 | 3 |
| 2013 | 2 |
| 2012 | 4 |
| 2011 | 4 |
| 2010 | 5 |
| 2009 | 2 |
| 2008 | 0 |
| 2007 | 1 |
| 2006 | 1 |
| 2005 | 0 |
| 2004 | 0 |
| 2003 | 1 |
| 2002 | 0 |
| 2001 | 0 |
| 2000 | 0 |
Use grade E
Fewer than 30 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 1
- Papers on reliability or validity
- 2
- Registered clinical studies using it as an outcome
- 0
Note on the counts: Original four-choice numeric facts; fixed item count; no effect if confidence or correction does not vary..
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 3 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Questions
- 24 original fact questions by default, or 16 drawn from the same bank. Each has four numeric answers and one correct answer, identical in all eight languages. Topics include shapes, chemistry, astronomy, time and musical instruments.
- First round
- Choose an answer without a deadline, then rate confidence from 1 (very unsure) to 6 (very sure). Two separate practice questions introduce the sequence.
- Feedback
- Every first answer is followed by the correct answer for 3 seconds by default, adjustable from 2 to 8 seconds. The chosen answer stays visible beside it.
- Delay
- 30 seconds of true-or-false arithmetic by default, adjustable from 0 to 120 seconds.
- Retest
- Answer every selected question again in a seeded new order. Answer positions are shuffled independently. No confidence rating or feedback is given during retest.
- Scores
- Goodman–Kruskal gamma between initial error confidence and correction at retest. Also reports correction rates for high confidence (4 to 6) and low confidence (1 to 3), their percentage-point difference, and all denominators.
This fixed-length, four-choice version uses our own questions. The original recall studies used typed answers and could collect a set number of errors. With no initial errors or no variation in confidence or correction, the association is undefined and stays blank. Each person takes this bank once; interpret the effect across a study and compare the same language and settings.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: hypercorrection_answer from the run with the most answer rows, linked by question ID to its confidence rows. Errors on the first round are the observations of interest, so low accuracy does not invalidate. Repeated choices or ratings do not invalidate. Answers under 200 ms on at least half of four or more questions invalidate. Missing pairs, fewer than six paired errors, an undefined gamma, or an earlier session are flagged for review. Hidden tabs withhold percentiles.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. Initial errors define the effect, so a knowledge gap is not a failed task. An association needs variation in confidence and correction. The six-error flag is a ConductScience choice. 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 |
|---|---|---|
| Questions | 24 questions (standard) | 16 questions, 24 questions (standard) |
| Correct answer display (seconds) | 3 | 2 to 8 |
| Delay before retest (seconds) | 30 | 0 to 120 |
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 Hypercorrection Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Butterfield & Metcalfe, 2001; Metcalfe & Finn, 2011), with the standard settings (questions: 24 questions; correct answer display (seconds): 3; delay before retest (seconds): 30). The main score was hypercorrection gamma (goodman-kruskal gamma between initial confidence and retest correctness, among initial errors with both observations).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Butterfield, B., & Metcalfe, J. (2001). Errors committed with high confidence are hypercorrected. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27(6), 1491-1494. https://doi.org/10.1037/0278-7393.27.6.1491Confidence in initial errors and their correction after feedback.
- Metcalfe, J., & Finn, B. (2011). People’s hypercorrection of high-confidence errors: Did they know it all along? Journal of Experimental Psychology: Learning, Memory, and Cognition, 37(2), 437-448. https://doi.org/10.1037/a0021962Question, confidence, feedback and retest procedure; this adaptation uses a fixed number of original multiple-choice facts.
Source papers
- Butterfield, B., & Metcalfe, J. (2001). Errors committed with high confidence are hypercorrected. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27(6), 1491-1494.Paradigm source, Adults
- Metcalfe, J., & Finn, B. (2011). People’s hypercorrection of high-confidence errors: Did they know it all along? Journal of Experimental Psychology: Learning, Memory, and Cognition, 37(2), 437-448.Paradigm source, Adults
Questions
How long does the Hypercorrection 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, tablets and phones. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Hypercorrection Gamma, High Minus Low Confidence Correction, High-Confidence Error Correction, Low-Confidence Error Correction, Error Correction Rate, Initial Accuracy, Retest Accuracy, Mean Confidence, Initial Errors, Paired Errors, High-Confidence Errors, Low-Confidence Errors, Corrected Errors, Paired Questions, First Answers, Retest Answers, Confidence Ratings, Planned Questions, Median Answer Time, Answers Under 100 ms and Earlier Sessions. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Hypercorrection Task used in research?
3 papers from 2021 to 2025 name it in their title or abstract. That ranks it 481 of 495 test families in the ConductCognition test census (use grade E). Source: ConductCognition test census 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Hypercorrection Task have?
In the published research, 1 paper reports 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?
Questions (16 questions, 24 questions (standard)), correct answer display (seconds) (2 to 8) and delay before retest (seconds) (0 to 120). 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, with its words in the same language. 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.