Triple-N dataset: large-scale fMRI-guided dense recordings of nonhuman primate neural responses to natural scenes

Authors: Yipeng Li, Xieyi Liu, Wanru Li, Jia Yang, Baoqi Gong, Wei Jin, Zhengxin Gong, Kesheng Wang, Jingqiu Luo, Zishuo Zhao, Pinglei Bao

Published: Nature Neuroscience, 10 June 2026 · Article type: Resource · Pages 1–13
DOI: 10.1038/s41593-026-02322-z

Abstract

Understanding high-level visual processing requires data that capture both fine-grained neuronal activity and large-scale cortical organization. We present the Triple-N dataset, which extends the Natural Scenes Dataset (NSD) framework to macaques by combining functional magnetic resonance imaging with dense Neuropixels recordings in the inferotemporal cortex and early visual areas during the viewing of 1,000 NSD images. Neuropixels probes provide high-resolution population sampling, capturing hundreds of simultaneously isolated units with millisecond temporal precision. Using these data, we show that inferotemporal category-selective regions exhibit robust tuning for their preferred categories, and dense sampling further reveals diverse temporal response patterns and image-dependent latency variations that reflect both intrinsic neuronal properties and stimulus features. Aligning macaque electrophysiology with human NSD functional magnetic resonance imaging demonstrates cross-species correspondences and divergences in representational geometry. Overall, the Triple-N dataset lays a foundation for unifying single-neuron dynamics, cortical representations and cross-species comparisons, helping to shape a more comprehensive understanding of primate visual processing.

Key Points

• Extension of NSD framework to macaques: 1,000 images from the Natural Scenes Dataset (originally human 7T fMRI, Allen et al. 2022) presented to macaques
• Dual-modality recording: fMRI (whole-brain, defining functional ROIs) + Neuropixels high-density electrophysiology (single-neuron resolution, millisecond precision)
• Recording targets: Inferotemporal (IT) cortex and early visual areas
• Population sampling: Hundreds of simultaneously isolated units per Neuropixels probe insertion
• Category-selective tuning: IT category-selective regions show robust tuning for preferred categories (faces, bodies, places, objects)
• Temporal dynamics: Diverse temporal response patterns and image-dependent latency variations reflecting intrinsic neuronal properties and stimulus features
• Cross-species alignment: Macaque ephys aligned with human NSD fMRI reveals both correspondences and divergences in representational geometry

Key Methods

• fMRI-guided targeting of Neuropixels probes to functionally defined regions of interest
• High-density silicon probe recordings (Neuropixels) providing simultaneous isolation of hundreds of units
• Presentation of 1,000 natural scene images from the NSD stimulus set
• Multi-level analysis: single-neuron, population, and fMRI-voxel resolution

References Cited

• Allen, E.J. et al. (2022). A massive 7T fMRI dataset to bridge cognitive neuroscience and artificial intelligence. Nat. Neurosci. 25, 116–126. 10.1038/s41593-021-00962-x — the original human NSD
• Jun, J.J. et al. (2017). Fully integrated silicon probes for high-density recording of neural activity. Nature 551, 232–236. 10.1038/nature24636 — Neuropixels probes
• Bao, P. et al. (2020). A map of object space in primate inferotemporal cortex. Nature 583, 103–108. 10.1038/s41586-020-2350-5 — IT cortex object selectivity
• Kriegeskorte, N. et al. (2008). Matching categorical object representations in inferior temporal cortex of man and monkey. Neuron 60, 1126–1141. 10.1016/j.neuron.2008.10.043 — cross-species representational comparison
• Vinken, K. et al. (2025). Mapping macaque to human cortex with natural scene responses. Proc. Natl Acad. Sci. USA 122, e2512619122. 10.1073/pnas.2512619122 — concurrent cross-species mapping work