Decision Making
Loss Aversion Task
Also called: mixed gambles task, gain-loss gambles, accept-reject gambling task, loss aversion coefficient
Measure how much more possible losses weigh than equal gains: accept or reject coin-flip gambles, scored as the loss aversion coefficient (lambda).
5-8 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
- 10
- Evidence grade
- C
- Registered clinical studies
- 4
What it measures
Measures how much more a possible loss weighs than an equal possible gain. 128 coin-flip gambles, each winning one amount of points or losing another, are accepted or rejected; a logistic fit gives the loss aversion coefficient (lambda).
Constructs
- Loss aversion
- Decision under risk
- Valuation of gains and losses
- Economic preferences
Research uses
- Behavioral economics and neuroeconomics research
- Decision-making in addiction, mood and anxiety research
- Risk preference across the lifespan
- Within-study group comparison
How the paradigm works
The loss aversion task measures how much more a possible loss weighs than an equal possible gain (Tom, Fox, Trepel and Poldrack, 2007). Each screen shows one gamble: a coin flip that wins one amount of points or loses another, with an equal chance of each. The participant accepts or rejects it, and rejecting keeps the points they have. The gains and losses vary independently over a grid written for this test: gains from 14 to 56 points in steps of 6 and losses from 7 to 28 in steps of 3, 64 gambles in all, each shown twice in a new order, 128 choices. As in the source, the gain range is twice the loss range, so a person who weighs a loss twice as much as a gain is indifferent on the grid's diagonal, where every gain is twice its loss.
More about the paradigm
The participant starts with 30 points. No outcome is shown during the task. At the end one choice is drawn at random and, if that gamble was accepted, a coin flip plays it, so every choice can count, as in the source, which paid real money for drawn decisions. One check question with feedback comes first, and there is no time limit. The study decides what the points are worth, and the payment line says whether they are exchanged for money.
The main score is the loss aversion coefficient, lambda. A logistic regression of accepting on the gain and the loss gives one slope for each, and lambda is the loss slope over the gain slope: 1 weighs a point lost like a point won, and 2 weighs it twice as much. A small ridge penalty keeps the fit finite when the choices split the grid perfectly. In the source, 16 adults had a median lambda of 1.93 (range 0.99 to 6.75) and were on average indifferent to gambles whose gain was twice the loss. The acceptance rate is reported second. Lambda depends on the ranges of gains and losses offered: median lambdas were 1.59 to 2.28 with a wider gain range, 1.00 to 1.23 with equal ranges and 0.81 to 1.01 with a wider loss range (Walasek and Stewart, 2015). Compare lambda within a study that uses the same grid. Answers under 100 ms are counted as anticipations and left out. Scores support research and do not diagnose.
Scores
- Loss Aversion (lambda)score
- Minus the loss slope over the gain slope of a logistic regression of accepting on the gain and the loss.
- Above 1 means a loss weighs more than an equal gain
- Acceptance Rate%
- Share of the answered gambles that were accepted.
- Informational
- Log Loss Aversionscore
- The natural log of lambda, for analyses that need a symmetric scale.
- Above 0 means a loss weighs more than an equal gain
- Gain Sensitivityscore
- Change in the log odds of accepting per point of gain.
- Informational
- Loss Sensitivityscore
- Change in the log odds of rejecting per point of loss.
- Informational
- Repeat Agreement%
- Share of the gambles answered the same way both times they were shown.
- Below 0.6 is flagged for review
- Check Correctflag
- 1 when the check question about rejecting a gamble was answered correctly, 0 when not.
- 1 supports interpretation
- Median Decision Timems
- Median time from a gamble's appearance to the answer.
- Informational
- Anticipationscount
- Answers under 100 ms after a gamble appeared, left out of the scores.
- Lower is better
- Choices Answeredtrials
- Answered gambles, out of 128.
- Informational
- Points Wonpoints
- Points from the drawn choice: the gamble's gain or minus its loss if it was accepted, 0 if it was rejected.
- Informational
Use and evidence
10 papers from 2021 to 2025 name this task in their title or abstract, down 47% compared with 2016 to 2020. That ranks it 370 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 1 |
| 2024 | 3 |
| 2023 | 2 |
| 2022 | 2 |
| 2021 | 2 |
| 2020 | 8 |
| 2019 | 3 |
| 2018 | 2 |
| 2017 | 3 |
| 2016 | 3 |
| 2015 | 3 |
| 2014 | 3 |
| 2013 | 1 |
| 2012 | 1 |
| 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 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
- 1
- Registered clinical studies using it as an outcome
- 4
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.
- Gambles
- 64, each a coin flip that wins 14 to 56 points (steps of 6) or loses 7 to 28 points (steps of 3), written for this test. As in the source, the gain range is twice the loss range.
- Order
- Each gamble twice, once in each of 2 rounds, in an order drawn from the session's seed: 128 choices, each after a 500 ms blank.
- Choice
- Accept or reject by clicking or tapping a button. No time limit and no outcome during the task.
- Practice
- One check question with feedback: rejecting a gamble keeps the points.
- Draw
- The participant starts with 30 points. At the end one choice drawn from the session's seed is played with a coin flip if its gamble was accepted.
- Points
- The study decides what points are worth, and the payment line says whether they are exchanged for money.
- Scores
- Lambda, the loss slope over the gain slope of a logistic regression of accepting on the gain and the loss, with a small ridge penalty on the slopes (1 per squared point). Answers under 100 ms are left out as anticipations.
- Checks
- No estimate of lambda, under 60% of gambles answered the same way both times and a wrong check answer are flagged and never exclude.
Tom and colleagues (2007) used a 16 by 16 grid of dollar gambles, a 3-second window and four answers from strongly accept to strongly reject. This test uses an 8 by 8 grid of its own points, each gamble twice, no time limit and two answers. Lambda depends on the ranges of gains and losses offered (Walasek and Stewart, 2015), so compare results within a study. 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: the scored run's gambles, 128 in a complete run. Fewer than half answered is invalid, and any other incomplete run is flagged. No answer is correct, so accuracy is not judged against chance, and accepting or rejecting every gamble does not invalidate. No estimate of loss aversion, the two showings of a gamble agreeing on under 60% of gambles and a wrong check answer are flagged for review and never exclude. Answers under 100 ms are counted as anticipations and left out of the scores, and answering in under 200 ms on at least half the gambles invalidates.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. Loss aversion is fitted to all the answers, and an answer under 100 ms after a gamble appears cannot reflect it. Someone answering at random gives the same answer to about half of the gambles shown twice, while one loss aversion with noisy choices rarely falls under 60%; an odd pattern can still be a real preference, so these checks flag and never exclude. 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 |
|---|---|---|
| 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 Loss Aversion Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Tom SM, Fox CR, Trepel C, Poldrack RA, 2007), with the standard settings (payment line: Hypothetical points). The main score was loss aversion (lambda; minus the loss slope over the gain slope of a logistic regression of accepting on the gain and the loss).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Tom, S. M., Fox, C. R., Trepel, C., & Poldrack, R. A. (2007). The neural basis of loss aversion in decision-making under risk. Science, 315(5811), 515-518. https://doi.org/10.1126/science.1134239
- Walasek, L., & Stewart, N. (2015). How to make loss aversion disappear and reverse: Tests of the decision by sampling origin of loss aversion. Journal of Experimental Psychology: General, 144(1), 7-11. https://doi.org/10.1037/xge0000039How lambda depends on the ranges of gains and losses offered.
Source papers
- Tom, Fox, Trepel and Poldrack (2007): the task, the logistic fit and a median lambda of 1.93 (range 0.99 to 6.75) in 16 adults. Walasek and Stewart (2015): how lambda moves with the ranges of gains and losses.Paradigm source, Adults
Questions
How long does the Loss Aversion Task take?
About 5 to 8 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?
Loss Aversion (lambda), Acceptance Rate, Log Loss Aversion, Gain Sensitivity, Loss Sensitivity, Repeat Agreement, Check Correct, Median Decision Time, Anticipations, Choices Answered and Points Won. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Loss Aversion Task used in research?
10 papers from 2021 to 2025 name it in their title or abstract, down 47% compared with 2016 to 2020. That ranks it 370 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 Loss Aversion Task have?
In the published research, 0 papers report its 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?
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.
Related tests