Battery guide
Decisions and Reward Battery guide
Four choice tasks for decision-making and reward research: delay discounting, reversal learning, card-deck choices and risk taking.
- Guide version
- 1.2.0
- Battery version
- 1.0.0
What it measures
Who it is for: Decision-making and reward research.
It measures Delay discounting, Reward learning and reversal, Decision-making under uncertainty and Risk taking.
Executive function: Planning, switching rules, handling conflict and making decisions.
Scores describe how a person performs on these tasks. Research use only.
Tests in order
The tests run in this order. A study started from the battery keeps the order and the settings.
1. Delay Discounting. Measures how much a delay lowers the value of money, using adjusting choices between an amount today and $100 later.
2. Probabilistic Reversal Learning. Measures reward learning and flexibility when the better of two symbols switches without warning.
3. Iowa Gambling Lite. Measures decision making under uncertainty with four card decks that differ in long-run gains and losses.
4. Balloon Risk. Measures risk taking as participants pump balloons for points and decide when to collect before a pop.
| Order | Test | Domain | Time |
|---|---|---|---|
| 1 | Delay Discounting | Executive function | 3 to 5 minutes |
| 2 | Probabilistic Reversal Learning | Executive function | 4 to 6 minutes |
| 3 | Iowa Gambling Lite | Executive function | 6 to 9 minutes |
| 4 | Balloon Risk | Executive function | 5 to 9 minutes |
Time and devices
Allow about 29 minutes for the whole battery. The table gives each test's time.
Participants can use a desktop or laptop, tablet or phone.
On a desktop or laptop the screen must be at least 1024 by 700 pixels.
A tablet or phone needs a screen of at least 360 by 640 pixels.
On phones and tablets, each test is compared only with sessions on the same kind of device. A test shows a percentile there once that reference group holds 50 valid sessions in the participant's age band; until then the report shows raw scores and says so.
Participants need Chrome, Edge, Firefox or Safari.
| Test | Timing | Devices |
|---|---|---|
| Delay Discounting | Untimed | Desktop or laptop, tablet or phone |
| Probabilistic Reversal Learning | Untimed | Desktop or laptop, tablet or phone |
| Iowa Gambling Lite | Untimed | Desktop or laptop, tablet or phone |
| Balloon Risk | Untimed | Desktop or laptop, tablet or phone |
Languages
Runs in all eight participant languages.
How to give it
Start a study from the battery in the app. The tests, their order and their settings are already set and stay fixed. If you change, reorder or customize a test, the study becomes a custom study and no longer carries the battery.
Each test opens with its instructions. Participants see these:
Delay Discounting: Choose between an amount of money today and $100 after a delay. The money is hypothetical.
Probabilistic Reversal Learning: Choose one of two symbols to win points. One symbol wins more often, and the better symbol can change.
Iowa Gambling Lite: Choose cards from four decks to win points. Some decks are worse than others over time.
Balloon Risk: Pump each balloon to add points, then collect before it pops. A popped balloon loses its points.
| Step | Before you invite participants |
|---|---|
| 1 | Tell participants that the money and points in the tasks are not paid out, unless your protocol pays them separately. |
| 2 | Ask participants to take the battery in one sitting, without breaks inside a task. |
| 3 | Pilot the battery yourself before you invite anyone. |
How each test is scored
The server scores each test when the participant finishes it. The primary score is the one the app compares with a norm. The other scores are reported with it.
Every result records its scoring version. For this battery they are Delay Discounting 2.0.0, Probabilistic Reversal Learning 2.0.0, Iowa Gambling Lite 2.0.0 and Balloon Risk 2.0.0.
| Test | Score | What it is | Better |
|---|---|---|---|
| Delay Discounting | Area Under the Curve (proportion, primary) | Area under the indifference points, with delays and amounts scaled from 0 to 1. A value of 1 means no discounting. | Higher means less discounting |
| Delay Discounting | Discount Rate (k) | Hyperbolic k fitted to the indifference points with V = A / (1 + kD), with the delay D in days. | Higher means steeper discounting |
| Delay Discounting | Indifference Point at 1 Month ($) | Amount today judged equal to $100 in 1 month. Points are also reported at 1 week, 6 months, 1 year and 5 years. | Higher means less discounting |
| Delay Discounting | Immediate Choice Rate (proportion) | Share of all choices that took the amount today. | Informational |
| Delay Discounting | Consistency Flag | 1 when the indifference points pass both Johnson and Bickel (2008) checks. 0 when a point rises more than $20 above the one before it, or the last point is not at least $10 below the first. | 1 supports interpretation |
| Probabilistic Reversal Learning | Perseverative Errors (primary) | After each reversal that followed a reached criterion, repeated choices of the previously better symbol before the first switch. The first trial after a reversal is not counted. | Lower is better |
| Probabilistic Reversal Learning | Trials to Criterion (trials) | Trials until 8 of the last 10 choices first picked the better symbol. | Lower means faster learning |
| Probabilistic Reversal Learning | Reversals Completed | Times the participant reached the learning criterion again after a reversal. | Higher is better |
| Probabilistic Reversal Learning | Reversal Errors | Choices of the worse symbol after the first reversal. | Lower is better |
| Probabilistic Reversal Learning | Win-Stay Rate (proportion) | Share of trials after a win on which the participant chose the same symbol again. | Informational |
| Probabilistic Reversal Learning | Lose-Shift Rate (proportion) | Share of trials after no win on which the participant switched symbols. | Informational |
| Iowa Gambling Lite | Net Score (primary) | Choices from decks C and D minus choices from decks A and B. | Higher is better |
| Iowa Gambling Lite | Net Score by Block | Net score within each block of 20 choices, reported as netScoreBlock1, netScoreBlock2 and so on. Rising block scores show learning. | Higher is better |
| Iowa Gambling Lite | Advantageous Choice Rate (proportion) | Share of choices from decks C and D. | Higher is better |
| Iowa Gambling Lite | Final Points (points) | Point total after the last card, starting from 2,000. | Informational |
| Balloon Risk | Adjusted Average Pumps (pumps, primary) | Mean pumps on balloons that did not pop. | Higher means more risk taking |
| Balloon Risk | Explosion Rate (proportion) | Share of balloons that popped. | Informational |
| Balloon Risk | Total Points (points) | Points collected across all balloons. | Informational |
| Balloon Risk | Risk Escalation (pumps per balloon) | Change in pumps from one balloon to the next, from a straight-line fit over balloons that did not pop. | Positive values mean rising risk taking |
| Balloon Risk | Adjusted Pumps by Block (pumps) | Adjusted average pumps within each block of 10 balloons, reported as adjustedPumpsBlock1, adjustedPumpsBlock2 and so on. | Informational |
Session quality rules
The app checks every session when it scores it and records one of four states.
An Invalid session stays in the study's results, labelled with its reason, and does not count toward norms.
A session can carry warnings and still be valid. The results list the warnings next to the scores.
| State | What it means |
|---|---|
| Norm-ready | The session passed every check and a norm exists for it, so the app can show a percentile. |
| Raw-only | The scores are usable, but the app shows no percentile. Typical reasons are no norm for the test, changed settings, or a warning that makes a comparison unfair, such as a restart or a screen too short for the display. |
| Invalid technical | The setup broke the timing or the task. Examples are leaving full screen, a browser tab that lost focus during a timed test, and long pauses. |
| Invalid behavioral | The responses suggest the task was not done as asked. Examples are a failed attention check, repeated switching away from the task, and accuracy at or below chance on a choice task. |
Published norms used
No published norm is used for any test in this battery.
ConductCognition also builds norms from its own participants. An age band of a test uses them once it holds at least 50 valid sessions on the same device and input type, run with standard settings and scored with the current scoring version. Sessions from paying customers' studies and from clinical or patient samples do not count.
A report shows a percentile only when the app has stored one for that session. Otherwise it shows the raw scores.
Every reference group covers ages 18 and over. No reference group covers a participant under 18, so their report shows raw scores only, and their sessions do not count toward norms.
How to cite
ConductScience, Inc. Decisions and Reward Battery guide, version 1.2.0. ConductCognition. https://conductcognition.com/batteries/decisions-reward/guide
Cite the guide version you used. Name the battery version (1.0.0) and the scoring versions listed in this guide.