Loading field map
For decades, mainstream economics treated the human brain as a black box. Preferences were revealed by choices, and the internal machinery that produced those choices was considered irrelevant to predicting market outcomes. Neuroeconomics emerged from a growing conviction that this black-box approach was a limitation, not a virtue. If economic models could be grounded in the actual neural mechanisms that generate decisions—emotion, learning, self-control, social cognition—then the field might explain not just what people choose but why they choose it, and when those choices will systematically deviate from rational benchmarks. The central tension that defines neuroeconomics is between the abstract, as-if rationality of neoclassical theory and the concrete, often messy biological reality of the decision-making brain.
The first major framework to bridge neuroscience and economic choice was the Somatic Marker Hypothesis, proposed by Antonio Damasio and colleagues in the early 1990s. Drawing on decades of clinical work with patients who had damage to the ventromedial prefrontal cortex (vmPFC), Damasio argued that emotion is not an obstacle to good decisions but a necessary component. Patients with vmPFC lesions showed normal scores on tests of logical reasoning yet made disastrous real-life choices—gambling away savings, staying in destructive relationships, failing to learn from punishment. The hypothesis proposed that bodily signals, or "somatic markers," generated by emotional responses to anticipated outcomes, are integrated in the vmPFC and guide decision-making by flagging options as good or bad before conscious deliberation begins.
This framework directly challenged the standard economic assumption that emotion is a source of error to be suppressed. Instead, it suggested that a person who cannot generate somatic markers is incapable of making advantageous choices, even with full information. The Somatic Marker Hypothesis introduced neuroeconomics to a core methodological tool: studying patients with focal brain lesions to infer causal roles for specific regions. It also set the stage for a lasting debate about whether emotion and cognition are separable systems or deeply integrated.
While the Somatic Marker Hypothesis focused on how the brain represents the value of options before a choice, the Reward Prediction Error Model, formalized by Wolfram Schultz, Peter Dayan, and Read Montague in 1997, addressed how the brain learns those values from experience. Recording from dopamine neurons in monkeys, Schultz discovered that these neurons do not fire simply when a reward arrives. They fire when the reward is better than expected and suppress firing when the reward is worse than expected. A fully predicted reward elicits no change. This signal—the difference between received and expected reward—is a reward prediction error (RPE).
The RPE model provided a precise computational mechanism for reinforcement learning, formalized in temporal-difference learning algorithms. It explained how organisms update the expected value of actions and stimuli over time, and it offered a neural basis for the kind of adaptive expectation formation that economists had long modeled abstractly. Unlike the Somatic Marker Hypothesis, which emphasized a single brain region and a relatively slow, feeling-based guidance system, the RPE model pointed to a fast, subcortical, and computationally tractable learning circuit centered on the midbrain dopamine system and the striatum. This framework transformed neuroeconomics by giving it a formal language—computational models of learning—that could be directly tested with both animal electrophysiology and human neuroimaging.
By the late 1990s, a third framework began to take shape, one that mapped the psychological distinction between intuitive and deliberative thinking onto specific neural circuits. The Dual-System Approach, most influentially articulated by Matthew Lieberman, Jonathan Cohen, and others, proposed that decision-making reflects the interplay of two broad neural systems: an automatic, emotional, "System 1" centered on the limbic system (including the amygdala and ventral striatum) and a controlled, deliberative, "System 2" centered on the prefrontal cortex (especially the dorsolateral prefrontal cortex, dlPFC).
This framework resonated strongly with behavioral economics, which had already documented systematic biases in judgment and choice that seemed to arise from fast, intuitive heuristics. The Dual-System Approach offered a neural explanation for those biases: when the intuitive system overrides the deliberative system, choices reflect emotional impulses rather than reasoned preferences. It also provided a neural account of self-control problems: successful self-regulation requires the dlPFC to inhibit limbic-driven urges. The framework coexisted with the RPE model by focusing on a different stage of decision-making—the competition between valuation systems at the moment of choice, rather than the learning process that built those valuations. However, it also generated a persistent tension: was the brain truly divided into two competing systems, or was decision-making better described by a single, integrated valuation network that flexibly recruited different regions depending on context?
As neuroimaging methods matured, researchers began to ask whether the behavioral models developed by Kahneman and Tversky had identifiable neural signatures. The Neural Correlates of Prospect Theory program, emerging around 2000, used fMRI to test whether the brain encodes the core elements of prospect theory—reference dependence, loss aversion, diminishing sensitivity—in ways that the standard expected utility model could not capture.
Key studies found that the amygdala and insula respond more strongly to potential losses than to equivalent gains, providing a neural basis for loss aversion. Activity in the ventral striatum tracked gains and losses relative to a reference point rather than absolute final wealth, consistent with reference dependence. The shape of the value function—steeper for losses than for gains—was reflected in differential neural responses in the vmPFC and striatum. This framework did not replace the earlier ones; rather, it absorbed and extended them. It used the Somatic Marker Hypothesis's emphasis on emotion and the Dual-System Approach's distinction between automatic and controlled processing, but it anchored those ideas to a specific, mathematically precise behavioral theory. The Neural Correlates program demonstrated that neuroeconomics could do more than discover new phenomena—it could test and refine existing economic models by examining whether their assumptions are neurally plausible.
The fifth framework, Social Neuroeconomics, broadened the field from individual choice to interactive decision-making. Starting around 2000, researchers began applying the tools of neuroeconomics—fMRI, computational modeling, lesion studies—to social behaviors such as trust, fairness, cooperation, and punishment. This framework drew heavily on the RPE model, showing that the same dopamine-driven learning signals that encode monetary rewards also encode social rewards like a fair offer or a cooperative partner. The striatum and vmPFC, central to the earlier frameworks, were found to represent the value of social outcomes in much the same way they represent the value of money or food.
Social Neuroeconomics also revived and transformed the Dual-System Approach by examining how social context modulates the competition between intuitive and deliberative systems. For example, rejecting an unfair offer in the Ultimatum Game—a costly punishment that standard theory predicts no rational agent would take—was linked to emotional responses in the anterior insula and to prefrontal control regions that override the impulse to accept any positive offer. This framework coexists with the Neural Correlates program by extending its methods to strategic interaction, and it complements the Somatic Marker Hypothesis by showing that social emotions like guilt and trust serve as somatic markers in interpersonal settings.
Today, all five frameworks remain active, but their influence and roles have shifted. The Reward Prediction Error Model has become the most formally rigorous and widely adopted framework, serving as the computational backbone for much of modern neuroeconomics. Its algorithms are used to model learning in everything from consumer choice to addiction to social interaction. The Dual-System Approach, while still influential in popular accounts, has been increasingly criticized as oversimplified. Many researchers now favor models in which a single valuation system, centered on the vmPFC and striatum, integrates multiple inputs (emotional, cognitive, social) rather than pitting two systems against each other.
The Somatic Marker Hypothesis remains important for its foundational demonstration that emotion is necessary for decision-making, but it has been partially absorbed into broader frameworks that treat somatic markers as one input among many into a unified valuation process. The Neural Correlates of Prospect Theory program has been largely successful in validating the neural plausibility of prospect theory, but it has also revealed complexities—such as individual differences in loss aversion that correlate with amygdala structure—that the original behavioral model did not predict. Social Neuroeconomics continues to expand, increasingly using computational models of social learning and causal methods like transcranial magnetic stimulation (TMS) to test whether neural activity in specific regions is causally necessary for social behavior.
What the leading frameworks agree on today is that decision-making is fundamentally a neural process that can be studied with the tools of neuroscience, and that economic models are improved by incorporating neural constraints. They disagree most sharply on the architecture of the decision system: is it best described as a competition between two qualitatively different systems (Dual-System), or as a single valuation system that flexibly weights multiple inputs (unified valuation)? This debate is unlikely to be resolved by neuroimaging alone; it increasingly requires causal interventions and computational models that can distinguish between competing architectures. The field's trajectory has moved from identifying neural correlates of economic behavior to building mechanistic, computationally explicit models that explain how the brain actually makes choices—a shift that promises to deepen the connection between economics and the rest of the life sciences.