Summary: Decisions begin subtly, as the brain accumulates evidence and leans toward a choice long before an action is taken. New research reveals that despite wide variability in individual neuron responses, a common underlying structure — a shared “potential landscape” — organizes neural activity and steers the brain toward coherent decisions.
Researchers trained rhesus macaques on a color-discrimination task while recording neural activity in the dorsal premotor cortex. Their analysis shows that individual neurons display diverse firing patterns, but these patterns are shaped by a unified dynamic framework that changes with task difficulty. The results provide a new model for how populations of neurons coordinate complex decisions and offer clues for understanding psychiatric conditions that impair decision processes.
Key Facts:
- Shared Structure: Neurons with varied responses follow a common potential landscape that organizes collective decision-making.
- Difficulty Modulation: Easier decisions correspond to steeper landscape slopes that favor rapid, decisive outcomes; harder decisions flatten the landscape and increase susceptibility to noise.
- Clinical Implications: Revealing how neurons coordinate choices may help explain decision-making disruptions in disorders such as schizophrenia and bipolar disorder.
Source: Princeton University
Every decision starts before we act.
Before any behavior appears, the brain is already sampling information, weighing alternatives, and gradually committing to a choice. Even when different people face the same evidence, their outcomes can differ—especially under uncertainty. For example, two drivers in traffic may perceive the same situation yet choose opposite responses: one accelerates to merge, the other brakes and waits.
How the brain—an immense network of specialized cells—reaches rapid, coordinated choices has been difficult to pin down. New collaborative work by researchers at Princeton University, Cold Spring Harbor Laboratory, Stanford University, and Boston University illuminates how diverse neurons cooperate to produce unified decisions.
The team trained monkeys to report whether red or green dominated a patterned display. Some trials were clear and easy; others were ambiguous and demanded more deliberation. While animals made choices, investigators recorded spikes from neurons in the dorsal premotor cortex, a region that transforms decisions into planned actions.
Individual neurons showed strikingly heterogeneous responses, even within the same trial, suggesting a complex and variable neural code for choices. Rather than concluding that this heterogeneity reflects inscrutable dynamics, the researchers developed a flexible computational model that explained diversity with two key elements: a neuron’s tuning—its preference for when and to which decision it responds—and population-level dynamics captured as a potential landscape.
In the model, the potential landscape shapes how population activity evolves. Valleys in this landscape correspond to stable decision states; activity rolls across the terrain like a ball, with steeper slopes driving faster convergence to a choice. When the landscape is flatter, noise can divert activity and increase the likelihood of errors.
Fitting the model to recorded spike data revealed that while single-neuron tuning remained consistent across easy and hard trials, the geometry of the potential landscape changed with task difficulty. Easy trials exhibited steep slopes that promoted rapid, confident choices. In ambiguous trials the landscape flattened, slowing the decision process and raising the chance of mistakes. Crucially, although each neuron showed its own unique firing pattern, all neurons appeared to be governed by the same underlying landscape.
“Imagine a group of skiers descending the same mountain,” said Tatiana Engel, Ph.D., associate professor at the Princeton Neuroscience Institute and senior author. “Each skier follows a slightly different path, but the slope beneath them shapes every trajectory. In the same way, individual neurons have distinct preferences, yet they are all guided by a shared landscape that draws the network to a stable decision.”
This unified coding principle links the dynamics of cognitive decision formation to geometric organization previously described for sensory representations. By revealing an attractor-like mechanism in the premotor cortex, the findings suggest a general framework for how populations of neurons encode both static sensory features and evolving cognitive variables during single trials.
Understanding how diverse neurons coordinate choices has implications beyond basic neuroscience. Disrupted decision dynamics are characteristic of several psychiatric disorders, and characterizing the geometry and dynamics that underlie healthy decision-making may help explain how these processes break down in disease.
With the model established, Engel and colleagues plan to investigate how different neuron types and their connectivity contribute to the varied tuning patterns and sequential phases of decision formation observed in single trials.
“Each decision is unique,” Engel said. “By examining single trials and single neurons, we can begin to reveal the principles that make choices reliable despite the diversity of individual neural responses.”
About this decision-making and neuroscience research news
Author: Daniel Vahaba
Source: Princeton University
Contact: Daniel Vahaba – Princeton University
Image: The image is credited to Neuroscience News
Original Research: Open access.
“The dynamics and geometry of choice in the premotor cortex” by Tatiana Engel et al. Nature
Abstract
The dynamics and geometry of choice in the premotor cortex
Sensory variables are represented by coordinated neural population activity, where single-neuron tuning curves define the geometry of population codes. Whether a similar coding principle applies to dynamic cognitive variables has been unclear because internal cognitive processes unfold with unique time courses on single trials, visible only in the irregular spiking of heterogeneous neurons.
Here, the authors demonstrate a population code for the dynamics of choice formation in the primate premotor cortex. They developed an approach to jointly infer population dynamics and the tuning functions of individual neurons to the population state. Applied to spike recordings during decision-making, the model revealed that neural populations encoded a common dynamic variable predicting choices, and that heterogeneous firing rates arose from diverse tuning of single neurons to that decision variable. The inferred dynamics are consistent with an attractor mechanism for decision computation, revealing a unifying geometric principle for neural encoding of both sensory and dynamic cognitive variables.