Neural coding of choice and outcome modulated by uncertainty in orbitofrontal vs motor cortex

• Neural coding of choice and outcome is modulated by uncertainty in orbitofrontal cortex but not in secondary motor cortex.¹
• Region-specific coding of uncertainty informs decoding and closed-loop BCI in non-motor domains.¹
• The finding is from Nature-level methods in computational neuroscience.^{1 1}

Gardner updates

• Neural coding of choice and outcome was shown to be modulated by uncertainty in orbitofrontal cortex but not in secondary motor cortex (Nature). ¹

Weekly enrichment (2026-07-20)

• The underlying study (Romero-Sosa et al., Nature Communications, 2025) imaged calcium activity (GCaMP6f via GRIN lens) from single neurons in rat orbitofrontal cortex (OFC, n = 8 rats) versus secondary motor cortex (M2, n = 6 rats) during de novo learning of increasingly uncertain reward-probability schedules.²³
• Binary support vector machine (SVM) decoders were trained on calcium traces to predict Chosen Side (left/right) and Trial Outcome (rewarded/unrewarded); choice decoding ramped up before trial initiation and peaked ~500 ms after the choice nosepoke.²
• M2 decoded choice with high accuracy under all certainty conditions, whereas OFC choice decoding rose to match M2 only under the least-certain schedule (Schedule 3: βArea = −0.0621, p = 0.073); M2 out-decoded OFC when contingencies were more certain (Schedule 1: p = 0.0027).²
• Behavioral strategies Win-Stay and Lose-Shift predicted decoding accuracy of choice and outcome in OFC but not in M2, linking OFC coding to adaptive strategy use under uncertainty.²⁴
• Chemogenetic (DREADD) inhibition of OFC neurons attenuated learning across all schedules, whereas M2 supported learning only in the most certain reward schedule, dissociating the two regions functionally.²⁵
• Ensemble size (number of neurons) was entered as a moderator and did not significantly explain decoding accuracy, and session/experience was not a significant predictor, arguing against trivial confounds.²
• OFC-implanted rats showed faster choice latencies than M2-implanted rats (βArea = −0.57, p < 0.01), and the M2 group exhibited attenuated learning relative to the OFC group (βarea, p = 0.0416).²
• BCI implication: uncertainty-dependent, region-specific coding suggests non-motor cortical areas such as OFC carry decodable choice/outcome signals useful for closed-loop decoding in volatile environments; this was demonstrated in rodents, not tested in humans here.²³

Footnotes

1. https://news.google.com/rss/articles/CBMiX0FVX3lxTE5wUnY2R3BIb2E1bnNCSXZyblAzMkdwWGVaeWZzaktCVS12dGMwU1BJbm83MGZrbncwdm9GQUs0RDlsaUV3VTBKZ1FUTkRTNHdyaU1SdWpHNFdKdjZhQTc4?oc=5 ↩ ↩² ↩³ ↩⁴ ↩⁵

2. https://www.nature.com/articles/s41467-025-63866-5 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸

3. https://pmc.ncbi.nlm.nih.gov/articles/PMC11580916/ ↩ ↩²

4. https://pubmed.ncbi.nlm.nih.gov/39574574/ ↩

5. https://www.biorxiv.org/content/10.1101/2024.11.05.622092v1 ↩