What is it about?
We make choices and learn from what happens next. Many of the brain systems that support these abilities sit deep inside the brain and are hard to study with standard scans. We used an ultra-strong MRI scanner (7 Tesla) and an analysis approach that looks at brain activity and behavior together at the same time, rather than in separate steps. We found that deep brain regions—especially parts of the striatum and thalamus—were tied to how urgently people moved toward a decision when speed was emphasized, more than to simple cautiousness. Several deep regions tracked how surprising the reward was on each trial, but, unexpectedly, classic dopamine midbrain areas did not; instead, the substantia nigra was more related to representing expected value. By analyzing brain and behavior jointly, we obtained stronger, less biased links than standard methods. These insights clarify how the human deep brain supports everyday decisions and learning and may inform conditions that affect these circuits, such as Parkinson’s disease and addiction.
Featured Image
Photo by Robina Weermeijer on Unsplash
Why is it important?
This work is unique in combining ultra-high-field 7‑Tesla MRI with individualized deep-brain maps to reliably measure tiny subcortical nuclei that are often invisible at standard field strengths. It also uses “joint modeling” to estimate brain activity and behavior together at the same time, avoiding the well-known bias from two-stage analyses. You empirically show how much standard methods understate brain–behavior links. It also makes an important theoretical contribution, as it shifts the interpretation of speed–accuracy control toward urgency mechanisms, refining how we think about decision-making systems in the human deep brain.
Perspectives
Two findings surprised me the most. First, the strong role of urgency (rather than caution) in deep-brain signals during speeded choices changed how I think about the speed–accuracy trade-off. Second, the absence of classic midbrain dopamine signatures for reward prediction errors—paired with robust value signals and reward prediction errors elsewhere—nudged me to reconsider the relationship between BOLD and neural firing patterns, which I think warrants deeper investigations.
Steven Miletic
Universiteit Leiden
Read the Original
This page is a summary of: Joint models reveal human subcortical underpinnings of choice and learning behavior, Proceedings of the National Academy of Sciences, September 2025, Proceedings of the National Academy of Sciences,
DOI: 10.1073/pnas.2502269122.
You can read the full text:
Contributors
The following have contributed to this page
