Decision Making
Paired Choices Task
Also called: narrow bracketing task, choice bracketing, concurrent decisions, dominated choices
Measure narrow bracketing: participants make two choices about points at once, then choose among the totals, scored by how often the pair they chose is beaten by one they turned down.
6-8 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
- 52
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures narrow bracketing, making choices one at a time. In 6 problems, participants make two decisions about points at once, each between an amount for sure and an even chance, where the pair most often chosen is beaten in every case by the opposite pair; the same problems then return as one choice among the four totals. The main score is the drop in beaten pairs chosen when the totals are shown.
Constructs
- Choice bracketing
- Decision under risk
- Dominance violations
- Risk attitude
Research uses
- Behavioral economics research
- Judgment and decision-making research
- Individual differences in decision making
- Within-study group comparison
How the paradigm works
The paired choices task measures narrow bracketing, making several choices one at a time rather than together (Read, Loewenstein and Rabin, 1999), in the concurrent decisions of Tversky and Kahneman (1981) and the examples of Rabin and Weizsäcker (2009). Participants start with 1,500 points. In each of 6 problems they make two decisions on one screen. The first is between an amount for sure and an even chance of two amounts, such as 200 points for sure or 460 points 50 times in 100 and 0 points otherwise. The second is between a sure loss and an even chance of a larger loss, such as 250 points taken away for sure or 500 points taken away 50 times in 100. Taking the sure amount in the first decision and the chance in the second, as caution with gains and risk taking with losses predict, gives a total that the opposite pair beats in every case: here 200 or -300 points against 210 or -250. In Tversky and Kahneman's hypothetical version, 73% of 150 respondents chose that pair and 3% the pair that beats it; when the two totals were shown side by side, all 86 respondents chose the better one. In Rabin and Weizsäcker's laboratory replication with real payments, 28% chose the beaten pair.
More about the paradigm
The 6 problems, with amounts written for this test, follow four shapes: Tversky and Kahneman's sure gain and sure loss, and Rabin and Weizsäcker's examples with a mixed chance against nothing, with every total above 0, and with a gain inside the risky loss. Every chance is 50 in 100, as in their survey's examples. In the standard order every problem comes first as two decisions on one screen, then again as one choice among the four totals of those decisions, each listed with its chances; the separate decisions come first because the totals show which pair is beaten. A setting runs one presentation alone, for a comparison between groups. The participant clicks or taps one option in each decision and confirms, and can change a pick until confirming; there is no time limit. One check question on adding and taking away points, with feedback, comes first. At the end one screen is picked at random and what was chosen there is played for points. The order, the side of the sure options, the order of the totals and the screen played come from the session's seed and are saved on the rows.
Scores
- Bracketing Effectscore
- Share of separate screens where a beaten pair was chosen minus the share of combined screens where a beaten total was chosen, from -1 to 1. Needs at least 2 screens of each presentation, so it stays empty when one presentation runs.
- Positive is narrow bracketing; 0 means showing the totals made no difference
- Beaten Pairs, Separate%
- Share of separate screens where the pair chosen was beaten in every case by another pair.
- Informational
- Beaten Totals, Combined%
- Share of combined screens where the total chosen was beaten in every case by another total.
- Near 0 expected; 0.5 or more is flagged for review
- Risky Choices, First Decision%
- On separate screens, the share of first decisions that took the even chance over the amount for sure.
- Informational
- Risky Choices, Second Decision%
- On separate screens, the share of second decisions that took the even chance over the sure loss.
- Informational
- Check Correctflag
- 1 when the check question was answered correctly, 0 when not.
- 1 supports interpretation
- Median Decision Timems
- Median time from a screen to Confirm, over screens at or after 100 ms.
- Informational
- Screens Under 100 mscount
- Screens confirmed less than 100 ms after they appeared.
- 0 expected
- Changed Choicescount
- Times a pick was changed before confirming.
- Informational
- Screenscount
- Screens scored.
- 12 with both presentations, 6 with one
- Planned Screenscount
- Screens the session planned.
- Informational
- Points Wonpoints
- Points the screen played at the end added or took away from the 1,500 starting points; negative is a loss.
- Informational
Use and evidence
52 papers from 2021 to 2025 name this task in their title or abstract, down 13% compared with 2016 to 2020. That ranks it 449 of 711 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 10 |
| 2024 | 17 |
| 2023 | 5 |
| 2022 | 11 |
| 2021 | 9 |
| 2020 | 10 |
| 2019 | 21 |
| 2018 | 11 |
| 2017 | 6 |
| 2016 | 12 |
| 2015 | 5 |
| 2014 | 2 |
| 2013 | 5 |
| 2012 | 4 |
| 2011 | 1 |
| 2010 | 4 |
| 2009 | 4 |
| 2008 | 6 |
| 2007 | 5 |
| 2006 | 2 |
| 2005 | 2 |
| 2004 | 1 |
| 2003 | 2 |
| 2002 | 1 |
| 2001 | 2 |
| 2000 | 0 |
Use grade D
30 to 99 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 0
- Papers on reliability or validity
- 2
- Registered clinical studies using it as an outcome
- 0
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 this task, not ConductCognition's version of it.
Source: ConductCognition test census 0.58.0, counted 13 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Problems
- 6 problems, each two decisions: an amount for sure or an even chance of two amounts, and a sure loss or an even chance of a larger loss. The amounts are written for this test, and every chance is 50 in 100. Participants start with 1,500 points.
- Beaten pair
- In every problem, the sure amount in the first decision with the chance in the second is beaten in every case by the opposite pair. For example, 200 points for sure with 500 points taken away 50 times in 100 totals 200 or -300 points, against 210 or -250 for 460 or 0 points with 250 taken away for sure.
- Presentations
- Both (standard): every problem first as two decisions on one screen, then as one choice among the four totals with their chances, 12 screens, separate screens first. One presentation: 6 screens of that kind only.
- Shapes
- Four shapes from the sources: a sure gain with a sure loss (2 problems), a mixed chance against nothing (2), every total above 0 (1) and a gain inside the risky loss (1).
- Choice
- Click or tap one option in each decision, then Confirm; picks can change until then. No time limit. One check question on adding and taking away points comes first. The order, the side of the sure options and the order of the totals are drawn from the session's seed.
- Points
- One screen is picked at random at the end, and what was chosen there is played for points. A setting adds the line that the study pays for points.
- Scores
- The bracketing effect: the share of beaten pairs chosen on separate screens minus the share of beaten totals chosen on combined screens. Also each share, and the share of risky choices in each decision.
Tversky and Kahneman (1981) and Rabin and Weizsäcker (2009) gave the separate and the combined versions to different groups. The standard order here gives both to each participant, separate first, and a setting runs one version for a comparison between groups. 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 screens, each a pair of choices made separately or one choice among the four totals. No choice is correct, so accuracy is not judged against chance, and choosing one kind of option throughout is accepted. A run with fewer than half its planned screens is invalid and any other shortfall is flagged. Screens confirmed under 200 ms on at least half the screens invalidate. A beaten total chosen on at least half the combined screens, or a wrong check answer, is flagged for review and never excludes.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. Choosing a beaten pair when the decisions are separate is the measure, not an error. When the totals are shown, the beaten one sits beside the option that beats it, and almost nobody chose it in the published combined presentations; choosing it often is usually read as inattention, but it is also part of the score, so it flags and never excludes. 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 |
|---|---|---|
| Presentations | Separate decisions, then totals (standard) | Separate decisions, then totals (standard), Separate decisions only, Totals only |
| 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 Paired Choices Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Tversky A, Kahneman D, 1981), with the standard settings (presentations: Separate decisions, then totals; payment line: Hypothetical points). The main score was bracketing effect (share of separate screens where a beaten pair was chosen minus the share of combined screens where a beaten total was chosen, from -1 to 1).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice. Science, 211(4481), 453-458. https://doi.org/10.1126/science.7455683Problems 3 and 4: two concurrent decisions, the pair most chose, and the combined totals that show it beaten.
- Read, D., Loewenstein, G., & Rabin, M. (1999). Choice bracketing. Journal of Risk and Uncertainty, 19(1-3), 171-197. https://doi.org/10.1023/A:1007879411489Narrow and broad bracketing of choices.
- Rabin, M., & Weizsäcker, G. (2009). Narrow bracketing and dominated choices. American Economic Review, 99(4), 1508-1543. https://doi.org/10.1257/aer.99.4.1508A laboratory replication with real payments, survey examples with even chances, and the four combined totals.
Source papers
- Tversky and Kahneman (1981): Problems 3 and 4, the concurrent decisions and their combined totals. Rabin and Weizsäcker (2009): a laboratory replication with real payments and survey examples with even chances, separate and combined.Paradigm source, Adults
Questions
How long does the Paired Choices Task take?
About 6 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?
Bracketing Effect, Beaten Pairs, Separate, Beaten Totals, Combined, Risky Choices, First Decision, Risky Choices, Second Decision, Check Correct, Median Decision Time, Screens Under 100 ms, Changed Choices, Screens, Planned Screens and Points Won. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Paired Choices Task used in research?
52 papers from 2021 to 2025 name it in their title or abstract, down 13% compared with 2016 to 2020. That ranks it 449 of 711 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.58.0, counted 13 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Paired Choices Task have?
In the published research, 0 papers report 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?
Presentations (Separate decisions, then totals (standard), Separate decisions only, Totals only) 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.