The Iowa Gambling Task: How Emotions Drive Decisions
Photo: Tima Miroshnichenko / Pexels
Four decks of cards. A stack of play money. And a patient who knows, intellectually, that two of the decks are rigged to ruin them — yet cannot stop choosing from those decks anyway. That paradox, first observed at the University of Iowa in the early 1990s, forced neuroscience to rethink something it had long taken for granted: that reason and emotion are separate systems, with reason in charge. The Iowa Gambling Task showed they are not.
What the Experiment Was Designed to Answer
The puzzle began with a specific type of patient. People with damage to the ventromedial prefrontal cortex (vmPFC) — a region of the frontal lobe roughly behind the forehead and above the eye sockets — presented a strange clinical profile. Their IQ scores were intact. They remembered facts, could reason through logic problems, and described the consequences of bad decisions eloquently. Yet in their actual lives, they made catastrophically poor choices: botched businesses, broken relationships, financial ruin. They knew what was good in the abstract but could not act on it.
Antonio Damasio, a neurologist at the University of Iowa College of Medicine, suspected the disconnect had something to do with emotion. He argued that the vmPFC damage was severing a link between the emotional "gut feeling" system and the cognitive machinery of planning and choice. To test that idea, he needed a controlled task that could simulate real-world decision-making under uncertainty in a laboratory setting.
The task he and his colleagues Antoine Bechara and Hanna Damasio developed would become one of the most-cited behavioral neuroscience paradigms of the past three decades.
Source: Bechara, A., Damasio, A. R., Damasio, H., & Anderson, S. W. (1994). Insensitivity to future consequences following damage to human prefrontal cortex. Cognition, 50(1–3), 7–15.
The Setup
The Iowa Gambling Task, as it became known, is deceptively simple. A participant sits at a table facing four decks of cards labeled A, B, C, and D. They start with a loan of $2,000 in play money. The instructions are minimal: turn cards from any deck you choose, one at a time, with the goal of making as much money as possible. You will sometimes win money. You will sometimes lose money. The task runs for 100 card selections.
What the participant does not know is that the four decks are rigged in a specific way.
Decks A and B are "bad" decks. They offer high immediate rewards — $100 per card — but also include occasional severe penalties that can reach $1,250 in a single turn. Played long enough, these decks produce a net loss.
Decks C and D are "good" decks. They offer modest rewards of $50 per card, and their penalties are smaller and less frequent. Over 100 cards, playing C and D produces a net gain.
The task is not announced as a test of strategy. There is no instruction to "figure out the decks." The point is to create an environment of genuine uncertainty — like real-world decisions — where participants must learn from accumulated feedback rather than from a rule they are handed.
Source: Iowa Gambling Task — Neurobs.com
The Protocol
The basic structure of a session is straightforward, but what Bechara and colleagues added transformed it into something far more powerful: they measured galvanic skin response (GSR) throughout the task.
Galvanic skin response is a measure of skin conductance — how easily an electrical signal passes across the surface of the skin. Sweat gland activity drives it. When you experience even a mild stress response or emotional activation, your sweat glands prime, skin conductance rises, and the GSR spike is detectable. It is one of the most sensitive, hard-to-fake physiological signals researchers have access to, which is why polygraph machines, however flawed as lie detectors, rely on it.
At regular intervals during the task, participants were also asked to verbally report their knowledge: Did they have a "hunch" about which decks were good or bad? Could they explain their strategy?
This dual recording — physiological signal plus verbal report — let the team track three things simultaneously as the task progressed:
- Which decks the participant was actually choosing from.
- Whether anticipatory GSR spikes appeared before choosing from the bad decks.
- When the participant became consciously aware that the bad decks were risky.
The participants were divided into two groups: healthy controls with no brain damage, and patients with vmPFC lesions confirmed by neuroimaging.
Source: Editorial: Twenty Years After the Iowa Gambling Task — PMC
The Results
The results split cleanly — and startlingly — along group lines.
In healthy controls, something happened around the 10th card selection that no one fully expected. Before participants could articulate any conscious understanding of why the decks differed, before they could explain that A and B were bad, their galvanic skin response began to spike anticipatorily when their hand hovered over decks A or B. The body was reacting to the risky deck before the mind had reasoned through why it was risky. By around the 50th card, most healthy participants had also developed conscious knowledge — they could explain, in rough terms, that C and D were better bets. Their behavior, already steering toward C and D, matched their emerging conscious understanding.
In vmPFC patients, neither thing happened. Their hands did not generate anticipatory GSR spikes over the bad decks. They did not shift their choices away from A and B with the same regularity as controls. Intriguingly, many patients could eventually verbalize that the bad decks seemed riskier — the declarative, fact-based knowledge was present — but it did not translate into changed behavior. Knowing and doing remained disconnected.
Think of it this way: a healthy player was like a thermostat that adjusts the temperature before you consciously decide you are cold, using feedback from the room. The vmPFC patient was like a thermostat with a broken sensor — the display might show the right number, but the heating system does not respond.
Source: Bechara, A., Damasio, H., Tranel, D., & Damasio, A. R. (1997). Deciding advantageously before knowing the advantageous strategy. Science, 275(5304), 1293–1295.
The Somatic Marker Hypothesis
The Iowa Gambling Task was never just a test of card preferences. It was designed to probe Antonio Damasio's broader theoretical framework, which he called the somatic marker hypothesis and laid out in his 1994 book Descartes' Error.
The central claim is this: the body generates physical signals — changes in heart rate, breathing, hormonal state, muscle tone — in response to anticipated outcomes. These signals (the "somatic markers") are registered in brain structures including the vmPFC and the amygdala, and they bias future decision-making toward or away from options associated with positive or negative past outcomes. The markers work faster than conscious reasoning, and they operate even when the reasoning has not yet caught up.
Damasio titled his book Descartes' Error deliberately. Descartes' famous mind-body dualism — the idea that the rational mind operates separately from and above the messy, physical body — had been absorbed into centuries of Western scientific thinking. The assumption was that good decisions require the suppression of emotion: the cooler the deliberation, the better the outcome. Damasio's patients, who had lost emotional input to their reasoning through vmPFC damage, gave the lie to that assumption. Stripped of emotional guidance, they did not become more rational. They became worse at deciding.
The IGT was the behavioral instrument that demonstrated this. The anticipatory GSR spikes in healthy participants showed that the body was "deciding" before the verbal system caught up — emotion was acting as a rapid-response guide, narrowing the option space before slow deliberation took over.
Source: Damasio, A. R. (1994). Descartes' Error: Emotion, Reason, and the Human Brain. Putnam.
The Timeline
| Year | Event |
|---|---|
| Early 1990s | Bechara, Damasio, and colleagues develop the card-deck paradigm at University of Iowa |
| 1994 | First formal publication in Cognition; Antonio Damasio publishes Descartes' Error |
| 1997 | Landmark Science paper: healthy participants generate anticipatory GSR before conscious knowledge |
| 2004 | Maia & McClelland challenge: participants may have more conscious knowledge than thought |
| 2005 | Bechara et al. publish direct response in Trends in Cognitive Sciences |
| 2010s | IGT adopted in clinical research for ADHD, addiction, and psychopathy populations |
| 2019 | Large-scale neuroimaging review confirms multiple brain systems involved in IGT performance |
The Controversy
No experiment with this level of influence escapes scrutiny, and the IGT generated substantial debate on two fronts: the methodology and the interpretation.
The Maia & McClelland challenge (2004) was the most pointed methodological critique. Bechara's group had used a simple verbal probe — "What do you know about the decks?" — to assess conscious awareness. Maia and McClelland argued this probe was too blunt. They redesigned the knowledge assessment to be far more sensitive, allowing participants to express degrees of confidence about each individual deck, and found that participants had much more detailed conscious knowledge than the original studies captured. Their conclusion was that healthy participants were not acting on nonconscious somatic signals — they were acting on explicit, albeit partially formed, strategic knowledge.
This was a direct challenge to one of the IGT's most striking claims: that the body "knows" before the mind. If conscious knowledge was present earlier than Bechara's probe detected, the sequence of events looked different.
Source: Maia, T. V., & McClelland, J. L. (2004). A reexamination of the evidence for the somatic marker hypothesis: what participants really know in the Iowa Gambling Task. PNAS, 101(45), 16075–16080.
Bechara's response in Trends in Cognitive Sciences (2005) acknowledged that conscious knowledge plays a role but maintained that the VMpFC patients' failure — and the dissociation between knowing and doing — still required the somatic marker framework to explain. Even if conscious knowledge arrives earlier than originally thought, the question remains: why can patients who explicitly state that decks A and B are risky still choose from them? The knowledge-action gap is real and reproducible.
A second, structural criticism concerns what "advantage" actually means in the task. Deck C, technically a "good" deck, produces many small wins but its loss schedule is spread across many cards. Some researchers have noted that participants sensitive to loss frequency (rather than magnitude) might rationally avoid deck C even if it is net positive. The task conflates different kinds of risk aversion, which complicates the claim that choosing A or B is simply "bad" decision-making.
Source: Is deck C an advantageous deck in the Iowa Gambling Task? — PMC
Replication and Reanalysis
The broad pattern of the IGT — vmPFC patients performing worse than controls — has replicated extensively across laboratories and cultures. A 2019 neuroimaging review published in SAGE Open confirmed that multiple brain regions are recruited during IGT performance, including the striatum, amygdala, and insula, alongside the vmPFC.
What has proven more fragile is the precise timing claim — that somatic signals precede conscious awareness — because this depends heavily on how "conscious awareness" is measured. When researchers use more sensitive probes, the window between anticipatory physiological response and reportable knowledge narrows considerably. The original 1997 claim that nonconscious body signals guide behavior before any explicit knowledge exists is now treated with more caution than it was in the late 1990s.
The task has also proven more complex to interpret in clinical populations than early researchers hoped. Performance on the IGT is influenced by working memory load, sensitivity to reward versus punishment, and learning rate — factors that vary across clinical groups. Patients with addiction, ADHD, or psychopathy may all perform poorly on the IGT, but for partially different underlying reasons.
Where It Stands Today
The Iowa Gambling Task remains one of the most widely used neuropsychological paradigms in research, with thousands of studies citing the original 1994 paper. Its value lies not in settling every theoretical question but in providing a standardized measure that can be compared across populations and conditions.
In ADHD research, IGT performance deficits are documented across multiple studies. Children and adults with ADHD tend to persist in selecting from disadvantageous decks longer than controls, suggesting difficulties in learning from feedback — consistent with the executive function deficits associated with the disorder.
In addiction research, individuals with substance use disorders consistently show impaired IGT performance, particularly hypersensitivity to immediate reward at the expense of long-term gain. Neuroimaging studies have found reduced gray matter volume in the vmPFC and orbitofrontal cortex of cocaine-dependent subjects, which correlates with their disadvantageous choices on the task.
In psychopathy research, the IGT is used to investigate whether psychopathic traits correlate with specific decision-making patterns, particularly reduced sensitivity to punishment signals — a prediction that follows naturally from the somatic marker framework, since psychopathy is associated with blunted physiological responses to threat.
In computational psychiatry, the IGT has become a benchmark task for reinforcement learning models of decision-making. By fitting learning parameters to individual IGT performance curves, researchers can extract estimates of how quickly a participant updates their expectations after wins and losses — a much richer picture than the binary "good deck / bad deck" classification the original task used.
The somatic marker hypothesis itself has evolved. Damasio and colleagues no longer claim that somatic signals are strictly necessary for all good decisions, or that they always precede conscious knowledge. The framework has been refined to a more nuanced position: somatic signals from the body and internal states contribute to decision-making, especially under uncertainty and time pressure, by biasing the system toward options with positive emotional histories. Emotion is not a distortion of reason; it is one of its inputs.
Source: Editorial: Iowa Gambling Task, Somatic Marker Hypothesis, and Neuroeconomics — PMC, 2022
FAQs
Q: What exactly are the four decks in the Iowa Gambling Task? A: Decks A and B are "bad" decks with high rewards ($100 per card) but frequent large penalties, producing a net loss over 100 trials. Decks C and D are "good" decks with moderate rewards ($50 per card) and smaller penalties, producing a net gain. Participants are not told which decks are which.
Q: What is the somatic marker hypothesis in plain language? A: The somatic marker hypothesis holds that the body generates physical signals — changes in heart rate, skin conductance, hormones — in response to imagined future outcomes. These signals are stored and accessed by brain regions including the vmPFC, and they guide decision-making by marking certain options as "feel bad" or "feel good" before deliberate reasoning finishes processing the situation.
Q: Why couldn't vmPFC patients just reason their way to the right answer? A: Some patients could verbalize that decks A and B seemed riskier — the declarative knowledge was present. But knowledge and action remained disconnected. The somatic marker system that translates emotional learning into behavioral bias was damaged, so the information did not reliably change what the patient chose. Knowing what to do and being guided to do it are handled by overlapping but distinct brain systems.
Q: Has the Iowa Gambling Task been replicated in other countries? A: Yes. The basic pattern — vmPFC patients performing worse than controls — has replicated across numerous laboratories in multiple countries. Cross-cultural performance differences exist in healthy participants (cultures differ in risk preference and reward sensitivity), but the vmPFC deficit holds across groups.
Q: Is the Iowa Gambling Task used in clinical diagnosis? A: It is used as a research tool and is sometimes included in neuropsychological batteries for research purposes, but it is not a diagnostic test on its own. Performance on the IGT can be affected by many variables — mood, working memory, anxiety — making it a research instrument rather than a definitive diagnostic measure.
Q: What is galvanic skin response and why does it matter for this experiment? A: Galvanic skin response (GSR) measures how easily an electrical current passes across the skin's surface. Sweat gland activity increases conductance even during mild emotional arousal. In the IGT, researchers measured GSR while participants contemplated each card choice. The critical finding was that GSR spiked anticipatorily — before selecting from a bad deck — in healthy participants, suggesting the body's stress system was flagging risky options before the verbal mind had caught up.
The Takeaway
The Iowa Gambling Task showed that the Enlightenment picture of reason-versus-emotion is wrong in an important direction. It is not that emotions corrupt rational decisions; it is that the absence of emotional input — through vmPFC damage — makes decisions worse, not better. The body keeps score on past experiences with risk and reward, and it votes on future choices through physiological signals that arrive before conscious deliberation finishes. The most precise version of this claim is still debated. But the basic finding — that feeling and thinking are not separate competitors but interlocked systems — has held through thirty years of replication and reanalysis.
Read next:
Weekly digest
Stay ahead of the curve
Get the best science, AI, and tech stories delivered to your inbox every week. No spam — unsubscribe any time.
About the Author
SciHub101 is an independent science and technology publication founded and edited by Buğra Sözeri, a software engineer and app developer based in Türkiye. Articles are drafted against peer-reviewed literature and primary sources, fact-checked, and edited by a human before publication. Where we compare products, anything built by our own team is labelled as such — see the editorial guidelines and About page for how corrections and disclosures work.
Related Articles
The Jenner Smallpox Vaccination: The First Vaccine
Edward Jenner vaccinated James Phipps with cowpox in 1796 and created the world's first vaccine. Learn how the experiment worked and why it ended smallpox.
Hubble's 1929 Redshift Observation: The Data That Proved the Universe Expands
In 1929, Edwin Hubble plotted galaxy distances against recession speeds and found a linear relationship. Here's what he measured, what it meant, and why the debate continues.
Why Do We Get Angry: The Science of Anger
How anger works through threat appraisal, the amygdala, and the fight-or-flight response. The neuroscience of anger, its functions in boundary-setting and motivation, why venting doesn't reduce anger, and evidence-based regulation strategies like reappraisal.
Why Music Affects Emotions: The Brain Science Explained
Why music affects emotions explained: dopamine and chills, prediction and expectation, why sad music feels good, earworms, and what neuroscience reveals about music's grip on the brain.
Why We Feel Emotions: The Science of Feeling
Why we feel emotions explained through James-Lange, Cannon-Bard, and Schachter-Singer theories, plus the amygdala's role and Lisa Feldman Barrett's constructed emotion model.
Why We Laugh: The Science of Humor and Laughter
Why we laugh explained: the neuroscience of humor, incongruity and relief theories, laughter as social bonding, why you can't tickle yourself, and what happens in the brain during a joke.