Control of Representation Updating by Higher-Order Thalamus Enables History-Based Decision-Making

AUTHOR INFORMATION

Patrick Steven Hosford^1,†, Hao Mei^1,†, Hosana Tagomori¹, Cillian Patrick Hayde¹, Manal Saeed Abdelaal¹, Hanna Tagomori¹, Miho Nakajima¹, Lukas Ian Schmitt^1,2,*

¹RIKEN Center for Brain Science, 2-1 Hirosawa, Wako-shi, Saitama, Japan.

²University of Tokyo, Department of Mathematical Informatics, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan

†These authors contributed equally

*Lead Contact. Corresponding Author Email: lukas.schmitt@riken.jp

MANUSCRIPT SUMMARY

When sensory information is ambiguous, we must interpret current inputs based on a representation constructed from our previous experiences. While appropriately updating this internal representation to track environmental changes is essential for behavior, the neural basis of this dynamic integration remains unknown. Using a combination of multi-area single-unit recordings and optogenetic techniques in mice both in behavioral and passive conditions, we find that interactions between the posterior parietal cortex (PPC) and its higher-order thalamic counterpart, the pulvinar (PUL) are necessary to stably maintain representations underlying decision-making based on sensory history. We also identify a mechanism by which shifts in statistical patterns across recent sensory experiences engage inhibitory control of the PUL by the thalamic reticular nucleus (TRN) to facilitate updating of encoded sensory history. Our results establish a framework in which complementary operations in the thalamus and cortex act in concert to allow internal representations to adaptively track changing conditions.

REPOSITORY INFORMATION

This repository contains supporting data and analysis/visualization code for findings included in the main figures of the manuscript (see ff_README for further information) as well as an implementation of the biologically inspired structured recurrent neural network developed as part of the research (see rnm_README for further information). Additional data collected for the study is available from the RIKEN Center for Brain Science neurodata repository system (https://neurodata.riken.jp/r/Schmitt/HosfordMeietalNeuron2026/).