Summary: New research identifies the orbitofrontal cortex (OFC) as a key brain region that enables animals to infer hidden states of their environment. In carefully controlled experiments with rats, scientists show that animals use inferred knowledge about reward availability to guide waiting behavior, and that silencing the OFC abolishes this capacity. These results clarify an essential cognitive mechanism and have implications for psychiatric conditions in which inference is impaired.
Researchers trained rats to wait for water rewards that varied across hidden, shifting conditions. Rats that learned the task adjusted how long they waited depending on whether the inferred state signaled a generally rich or poor reward environment. When the OFC was inactivated, trained animals no longer adjusted waiting times based on inferred states, demonstrating the OFC’s central role in building and updating internal models of the world.
Key Facts:
- OFC role: The orbitofrontal cortex supports state inference by updating beliefs about hidden environmental conditions.
- Behavioral evidence: Trained rats waited longer for the same water amount when it came from a port associated with a low-reward state versus a high-reward state, indicating they inferred overall reward context.
- Disruption effect: Temporarily disabling OFC activity removed these state-dependent differences in waiting behavior, confirming the region’s necessity for flexible inference-based decisions.
Source: NYU
Background: Animals survive in changing and uncertain environments not only by reacting to immediate cues but by forming inferences—internal assessments about hidden or latent states that affect future outcomes. For example, an animal might learn that a particular sound does not indicate danger and therefore should not trigger escape behavior later. How the brain constructs and uses these hidden-state inferences has been a central question in neuroscience.
A team at New York University reports in Neuron that the orbitofrontal cortex functions as an “inference engine,” updating beliefs about hidden states as environmental evidence accrues. The work combines behavioral testing, electrophysiology, neural population analyses, and targeted inactivation to reveal how OFC dynamics reflect and drive state inference.
“To survive, animals cannot simply react to each stimulus; they must generalize and make inferences,” says Christine Constantinople, professor at NYU’s Center for Neural Science and the study’s senior author. “This study pinpoints a specific computation performed by the OFC and shows how that computation supports adaptive behavior.”
The experimental task was a temporal wagering paradigm in which rats decided how long to wait for a water reward. Rats experienced discrete water volumes (5–80 microliters) delivered at ports. The distribution of volumes across trials defined hidden reward “states” labeled low, high, or mixed. Crucially, individual reward amounts could appear in more than one state—for example, a 20-microliter delivery occurred in low, mixed, and high states—so the true state was not directly observable on a single trial.
Well-trained rats used trial history to infer the current state and adjusted waiting time accordingly: they waited longer for a given water amount when it occurred in a low-reward state because the expectation of future rewards was lower, making waiting worthwhile. The same amount in a high-reward state was worth less relative to available alternatives, so rats waited for a shorter time. Untrained rats did not show this state-sensitive pattern.
To test causality, researchers disrupted OFC activity and found trained rats could no longer update beliefs about hidden states or modulate waiting based on inferred context. Electrophysiological recordings from over 10,000 neurons and novel population analyses revealed latent neural factors in OFC activity that tracked inferred states. These neural signals changed abruptly after trials that provided information about state transitions, supporting the idea that OFC dynamics encode belief updates used for inference.
Together, the behavioral, causal, and neural evidence indicate the OFC implements a precise computation: updating beliefs about latent states to guide flexible decision-making. These findings offer a mechanistic link between neural population dynamics and inference-based behavior and provide a framework for understanding conditions in which inference breaks down, including schizophrenia and bipolar disorder.
Funding: This work was supported by grants from the National Institutes of Health (DP2MH126376), the National Science Foundation (R01MH125571, K01MH132043, 5T32MH019524, 5T90DA043219, and F31MH130121), and the Department of Energy. The content is the authors’ responsibility and does not necessarily represent official NIH views.
Key Questions Answered:
A: The orbitofrontal cortex (OFC) supports updating beliefs about changing, hidden reward states and enables inference-based decisions.
A: Rats performed a temporal wagering task, choosing how long to wait for water rewards whose overall availability varied across hidden states, revealing whether rats inferred richer or poorer environments.
A: Disrupting OFC activity prevented trained rats from updating expectations about hidden states and eliminated state-dependent differences in waiting behavior, indicating the OFC is necessary for inference.
Editorial Notes:
– Article prepared by a Neuroscience News editor.
– Underlying journal paper reviewed in full.
– Additional explanatory context provided by staff.
About this research news
Author: James Devitt
Source: NYU
Contact: James Devitt – NYU
Image credit: Neuroscience News
Original research (open access): “The orbitofrontal cortex updates beliefs for state inference” by Christine Constantinople et al., Neuron.
Abstract
The orbitofrontal cortex updates beliefs for state inference
Although the orbitofrontal cortex (OFC) has been linked to learning and inferring latent states, the specific computation it performs for state inference remained unclear. This study shows that rat OFC activity updates beliefs about hidden states and that these belief updates are evident in OFC dynamics only when animals use an inference strategy rather than simpler alternatives.
Rats were trained on a temporal wagering task with hidden reward states. Well-trained animals used state inference to decide how long to wait for rewards, and OFC inactivations impaired belief updating. Electrophysiology and novel population analyses identified latent neural factors in OFC activity that reflected inferred states; these factors and firing rates exhibited abrupt changes after trials that provided information about state transitions. These results specify a computation carried out by OFC and reveal neural signatures of inference.