Why do people make different choices when faced with the same options? Standard models of decision-making explain variability in terms of noisy evidence accumulation or shifting attention. But a study published July 4 in Nature Communications suggests a deeper source of variability: the spontaneous, sub-second fluctuations of neural activity that occur before we even see the choice.
Using a closed-loop intracranial brain-computer interface, researchers at Grenoble Institute of Neuroscience and the Paris Brain Institute demonstrated that transient bursts of high-frequency activity in the anterior insula, a region long implicated in emotional awareness and interoception, directly bias whether a person accepts or rejects a complex offer. The effect occurs within milliseconds, far faster than conscious deliberation.
“We developed a closed-loop BCI that detects spontaneous fluctuations in broadband gamma activity and uses them to trigger stimulus presentation,” said corresponding author Julien Bastin of Grenoble Institute of Neuroscience. “This allowed us to ask whether the brain state at the moment an offer appears influences the decision, without the participant having any control over the timing.”
How the experiment worked
The team implanted intracranial electrodes in patients undergoing monitoring for epilepsy, allowing them to record local field potentials directly from the anterior insula with millisecond precision. They designed a closed-loop system that monitored broadband gamma activity (70-150 Hz) in real time. When activity crossed a pre-set threshold, either high or low, the system presented a multi-attribute offer combining pleasant and unpleasant components. Participants then decided whether to accept or reject the hypothetical offer.
The critical variable was whether the offer appeared during a moment of high or low pre-offer neural activity in the anterior insula. The participants themselves had no awareness of the timing manipulation.
The results were striking. Offers preceded by high broadband gamma activity in the anterior insula were significantly more likely to be accepted, even when the offer contained an unpleasant component that participants would normally reject. The high pre-offer activity was followed by a transient suppression of the same signal after the offer appeared, suggesting that the insula’s response to the offer was modulated by the state it was in just before.
“This challenges current neuro-computational models that treat each decision as if the brain starts from a neutral baseline,” said first author Clarissa Baratin. “We show that intrinsic brain states, spontaneous fluctuations on a sub-second timescale, directly shape choice behavior.”
Why the anterior insula
The anterior insula is a cortical region that integrates signals from the body, heartbeat, respiration, visceral sensations, with emotional and cognitive information. It has been consistently implicated in interoceptive awareness (the sense of the body’s internal state) and in processing negative emotions like disgust, pain, and unfairness. But its role in value-based decision-making has been debated: is it computing the value of options, signaling anticipation, or mediating the influence of body state on choices?
The new results support a specific role: the anterior insula’s spontaneous activity level sets a “gain” or “bias” that influences how subsequent choice information is processed. High pre-offer activity makes the brain more receptive to accepting offers that mix pleasant and unpleasant elements, perhaps because the insula’s interoceptive signal momentarily shifts the balance between approach and avoidance.
“This is a very different role from computing the value of options,” said co-author Mathias Pessiglione of the Paris Brain Institute, an expert in decision neuroscience. “The insula may not be telling the brain how good or bad something is. It may be changing the overall disposition to accept or reject based on ongoing internal states.”
Implications for decision neuroscience
The findings add to a growing recognition that spontaneous neural activity, long dismissed as noise in task-based fMRI and electrophysiology studies, carries meaningful information and exerts causal influence on behavior. For decision-making models, this poses a fundamental challenge. Standard frameworks like drift-diffusion models assume that decisions start from a neutral state and accumulate evidence over time. But if the starting state varies from trial to trial based on sub-second fluctuations in a dozen brain regions simultaneously, the models need to account for initial conditions that are anything but neutral.
The closed-loop BCI approach itself is also notable. Rather than passively recording and correlating brain activity with behavior post-hoc, the system actively probed causality: it used brain activity to gate stimulus presentation, turning a correlational observation into a causal manipulation. This approach could be extended to other brain regions and cognitive domains, mapping the causal influence of ongoing neural dynamics on perception, memory, and motor control.
Caveats and limitations
The study was conducted in patients with epilepsy who had electrodes implanted for clinical monitoring, which raises questions about generalizability, though the anterior insula recordings were from healthy tissue not involved in seizure onset. The task involved hypothetical multi-attribute choices, not real rewards or punishments, and the effect size, while statistically significant, was modest. The closed-loop system detected only anterior insula activity, leaving open the question of how interactions with other regions, the prefrontal cortex, striatum, amygdala, shape the full decision process.
Still, the finding that a sub-second fluctuation in one brain region can measurably shift a decision represents a clear challenge to models that treat the brain as a decision-making machine starting fresh at each trial.
Source: Baratin C, Pessiglione M, Kahane P, Robin A, Minotti L, Becq GJPC, Bastin J. Closed-loop readout of anterior insula high-gamma activity steers value-based decisions. Nature Communications (2026). DOI: 10.1038/s41467-026-75265-5