The Information Sampling Task: How Much Evidence Is Enough?
The Information Sampling Task (IST) is a computerized paradigm that asks participants to decide when they have collected “enough” information before committing to a choice—quantifying the trade-off between evidence, time, and reward. Popularised by Clark and colleagues (2006) to probe reflection–impulsivity in current and former substance users, the IST has since become a staple in decision-neuroscience, psychiatry, and neuro-economics.
Table of Contents
- Conceptual Roots
- Canonical Task Designs
- Key Dependent Measures
- Neural Correlates & Computational Models
- Clinical and Developmental Findings
- Design & Implementation Tips
- Common Pitfalls
- Key Takeaways
Conceptual Roots
The IST was built to capture evidence-accumulation under explicit sampling cost, bridging two literatures:
- Beads / Jumping-to-Conclusions Task – draws until confident which jar beads come from.
- Sequential-sampling models – drift-diffusion, Bayesian ideal observer.
By penalising premature or over-extended sampling, IST situates behaviour on a reflection ↔ impulsivity continuum—useful for disorders where decision latency matters.
Canonical Task Designs
| Variant | Mechanics | Typical Incentives |
|---|---|---|
| Fixed Win (FW) | 5×5 grid of grey boxes; click to reveal colour; all correct answers pay +100; sampling free but costs time. | |
| Decreasing Win (DW) | Starts at +250; each click reduces stake by −10; wrong answer loses remaining stake. | |
| Cost-Per-Sample (CPS) | Monetary penalty deducted per reveal; isolates cost sensitivity. | |
| Social-Information Variant | Participants view peer choices instead of colours; assesses social sampling. |
Task ends when participant presses “decide” and identifies the majority colour (usually Blue vs. Yellow).
Key Dependent Measures
- Draws to Decision (DTD) – number of samples before choice; lower DTD = impulsive.
- Probability of Correct Decision – Bayesian probability given evidence at decision-point.
- Sampling Efficiency – net points earned vs. optimal Bayesian agent.
- Decision Latency – ms between final sample and choice; reflects motor impulsivity.
Neural Correlates & Computational Models
- Frontopolar & dorsolateral prefrontal cortex (dlPFC) track cumulative evidence and prospective value of further sampling.
- Anterior cingulate cortex (ACC) signals conflict/cost trade-offs; theta-band EEG power predicts additional draws.
- Basal ganglia & striatum encode “urgency” signals that speed commitment as time-on-task grows.
- Bayesian and Drift-Diffusion fits show steeper urgency gain (k) in impulsive cohorts, replicating across IST and bead tasks.
Clinical and Developmental Findings
| Population | Pattern |
|---|---|
| ADHD | Fewer draws, lower sampling efficiency; symptom severity correlates with DTD. |
| Substance-Use Disorders | Cocaine and alcohol users jump-to-conclusions in DW variant; related to altered reward discounting. |
| Obsessive-Compulsive Disorder | More draws than controls—“over-checking” reflection extremes. |
| Adolescence | U-shaped curve: early teens impulsive, late teens optimise, elders cautious. |
| Parkinson’s (ON vs. OFF meds) | Dopamine agonists decrease draws, mirroring medication-related impulsivity. |
Design & Implementation Tips
- Calibrate Sampling Costs – pilot to avoid ceiling (everyone samples maximally) or floor (everyone guesses early).
- Randomize Majority Colour & Box Positions – prevents pattern learning.
- Collect Reaction Times – micro-second precision enables drift-diffusion modelling.
- Include Practice Trials with Feedback – teaches mechanics without leaking optimal strategy.
- Log Sample Sequence – trajectory data reveal changes-of-mind dynamics.
Open-source code exists in PsychoPy, jsPsych, and MATLAB/Python (e.g., online-IST GitHub repo).
Common Pitfalls
| Issue | Mitigation |
|---|---|
| Understanding Failure | Provide visual countdown of remaining stake; comprehension quiz. |
| Motor Fatigue | Limit trials to 40–60; interleave breaks. |
| Ceiling Accuracy | Use near-equal base-rate colours (e.g., 12 Blue, 13 Yellow) for harder inference. |
| Confound with Working Memory | Keep grid visible; avoid hidden memory component unless deliberately manipulating load. |
Key Takeaways
- The Information Sampling Task quantifies how much evidence people require before deciding, indexing reflection–impulsivity and cost-benefit calibration.
- Draws-to-decision and sampling efficiency map onto frontal-striatal circuits and distinguish clinical groups such as ADHD, OCD, and addiction.
- Careful design of sampling costs, trial difficulty, and feedback is crucial for interpretable data.
- IST variants are now used in pharmacology, lifespan development, and social decision-making, making the paradigm a versatile probe of cognitive control and uncertainty management.
Try it Now
Start the Information Sampling Task on kuakua.app
References
- Clark, L., Robbins, T. W., Ersche, K. D., & Sahakian, B. J. (2006). Reflection impulsivity in current and former substance users. Biological Psychiatry, 60(5), 515-522.