Sign in

Samuel Le Meur-Diebolt

@sdiebolt.bsky.social
87 followers 112 following 44 posts

Research Fellow in CortexLab‬ @UCL 🧑‍🔬 • Studying macroscale brain activity using fUSI in mice 🐭🧠

PostsRepliesMedia
Reposted by Samuel Le Meur-Diebolt
Emilie Macé @mace-lab.bsky.social · 01/10/2026
Exciting times for functional ultrasound! The field is growing fast—and we need to train the next generation. We’re launching the first Paris Ultrasound Workshop, on Dec 8–9 2026. 🎓 Leading experts 🧰 Open-source tools 👩‍🔬 For new & experienced users Register by Nov 15! Link: tinyurl.com/ParisfUSI
185
Reposted by Samuel Le Meur-Diebolt
Patrick Mineault @patrickmineault.bsky.social · 30/09/2026
There's a new imaging modality in town: functional ultrasound imaging. I went down a deep rabbit hole to understand how it works physically and computationally. Follow me as I show you how *you* can use it and improve it www.neuroai.science/p/functional...
neuroai.science
Functional ultrasound imaging from scratch
A new modality is coming to brain imaging
073
Reposted by Samuel Le Meur-Diebolt
napari @napari.org · 26/09/2026
🚀 napari is hiring Remote Developers-in-Residence! Join us to build the future of interactive n-dimensional data visualization in Python. 💻 60% Dev / 20% Maintenance / 20% Community & Outreach ⏱️ Contract, 14–35 hrs/wk 🌍 Global 100% Remote Learn more and apply today: napari.org/island-dispa...
napari.org
We’re hiring Developers-in-Residence! — napari blog documentation
Organization: napari (a NumFOCUS Sponsored Project) Position Type: Independent contractor agreement. Global, 100% Remote. Duration: One year, possibility to extend. Commitment: 0.4 to 1.0 FTE…
01116
Reposted by Samuel Le Meur-Diebolt
The Royal Society @royalsociety.org · 29/08/2026
The Ferrier Medal and Lecture 2027 is awarded to Professor Kenneth Daniel Harris for his contributions to neuroscience by unifying experimental neurophysiology with advanced analytical methods and for pioneering methods that allow recording from thousands of neurons. #RSMedals bit.ly/4qCJZRo
A man in a black shirt smiles at the camera in front of a whiteboard covered in scientific notation.
47219
Reposted by Samuel Le Meur-Diebolt
Matteo Carandini @carandinilab.net · 17/08/2026
I am proud of this paper. Since it came out in 2022 it has received 0 citations (yes that's a zero). But multiple young scientists have found it useful! (from other labs, so they were not coerced...). Perhaps you will find it useful too? SOME TIPS FOR WRITING SCIENCE doi.org/10.1523/ENEU...
doi.org
Some Tips for Writing Science
When preparing a scientific paper, we typically write with coauthors who have different backgrounds and styles, and we target readers who have little time and patience. To help both readers and writer...
616554
Reposted by Samuel Le Meur-Diebolt
SWC/GCNU Neuroinformatics Unit @neuroinformatics.dev · 17/08/2026
The Neuroinformatics Unit Open Software Summer School starts today! Week one starts with: - Animals in Motion, how to use open source tools to quantify animal behaviour - Large Array Data, working with very large scientific datasets neuroinformatics.dev/open-softwar... #NIU_OSSS_26
163
Reposted by Samuel Le Meur-Diebolt
Anyi Liu @anyiliu.bsky.social · 29/07/2026
Pyramidal cells in the cortex send a long apical dendrite to layer 1. What is its role? What happens when it's gone? ✂️ A causal spatial role for apical dendrites in the cortex Out today with @kenneth-harris.bsky.social, @carandinilab.net and @fedrossi.bsky.social www.biorxiv.org/content/10.6...
biorxiv.org
A causal spatial role for apical dendrites in the cortex
Cortical pyramidal neurons project a long apical dendrite to layer 1 (L1), to receive synaptic inputs that are thought to provide contextual and possibly suppressive modulation. In visual cortex, thes...
03215
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 30/07/2026
Hello, #fUSI users! We’re running a poll to rename the structural ultrasound imaging datatype in the fUSI-BIDS specification so it better reflects uses beyond angiography (e.g., including B-mode imaging). Join the fUSI Discord and share your opinion: discord.gg/5HFTcukKFt
041
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 21/07/2026
@dasasgue.bsky.social @jcmariani.bsky.social @chiarapepe.bsky.social you might have a better understanding of this than me! I’d be really grateful for your opinion on this 🤩
011
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 20/07/2026
Here's a quick Python script for the toy example: github.com/sdiebolt/cap... P.S. I'd guess the <0.1 Hz low-pass filter typically used in fMRI and fUSI has a similar impact!
000
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 20/07/2026
Would be really curious whether this is already obvious in the CAP literature, and how people usually think about or rule out this interpretation in practice!
100
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 20/07/2026
And in this toy example, the recovered CAPs tracks amplitude > mean better than true activation. Here's one example run: - CAP vs. "true active": 0.765 - CAP vs. "ROI amplitude > ROI mean": 0.937
100
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 20/07/2026
Running CAP clustering with k=2 (spherical k-means + sample-wise z-scoring, which amounts to correlation-based clustering), we recover a CAP/anti-CAP pair.
100
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 20/07/2026
I made a tiny toy simulation to illustrate this intuition: - one square ROI - stochastic positive-only activity, with explicit off-periods - voxelwise z-scoring
100
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 20/07/2026
So in principle, one latent positively activated network might get split into: - a CAP for higher-amplitude activations, - an anti-CAP for lower-amplitude / below-mean periods, even if the underlying process is not a real activation/deactivation toggle.
100
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 20/07/2026
If a network has stochastic positive activation above baseline, then z-scoring centers each voxel around its mean. So some genuinely active frames will end up positive, and some still-active frames will end up negative.
100
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 20/07/2026
But CAP analysis is generally performed after temporal z-scoring. That made me wonder: could at least some apparent CAP/anti-CAP structure come from z-scoring itself?
100
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 20/07/2026
Hi co-activation patterns people! I've been playing around with CAPs recently and wanted to sanity-check something. My understanding is that CAP vs. anti-CAP is usually interpreted not as "activation vs. no activation", but as two opposing configurations (two poles of the same fluctuation).
211
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 18/07/2026
Absolutely! I’ll fill out the survey. I plan on releasing the plugin in about a month, just waiting for the backend software that handles the heavy image processing to be released too. Thanks for the invitation!
010
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 17/07/2026
Developing a GUI for #fUSI acquisition has been so easy thanks to napari’s plugin system. Especially with today’s coding agents, having extensible platforms like napari to build on makes creating tools for niche scientific applications feel like a breeze!
010
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 17/07/2026
napari is such an incredible piece of scientific software! 🚀 And it's not just for visualizing static data. I've been using it for the past 6 months to drive a @verasonicsinc.bsky.social ultrasound system for #fUSI, and get a nice GUI for sequence control and live visualization! 🤩
293
Reposted by Samuel Le Meur-Diebolt
napari @napari.org · 12/07/2026
Struggling for something to do in a TWO DAY gap without ⚽? How about trying our pre-release of napari 0.8.0? napari.org/dev/release/... Of note, this release drops Py3.10 support and deprecates PyQt5 in favor of PyQt6. It also adds some neat new features and improvements! See below: 👇
1259
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 09/07/2026
Remember folks: you can ship some of the most advanced LLMs on Earth, but choosing distinguishable colors for a plot remains an open research problem. openai.com/index/previe...
031
Reposted by Samuel Le Meur-Diebolt
Matteo Carandini @carandinilab.net · 10/06/2026
Losing one side of Superior Colliculus is bad, but losing both sides seems fine. Wait, what? Flora & team discovered why: both sides press for contra actions and against inaction, obeying an incredibly simple additive rule. Out today in bioRxiv: www.biorxiv.org/cgi/content/...
A diagram showing a sequence of linear operations, with visual signals from visual cortex added to auditory signals from auditory cortex. Prefrontal cortex sums these signals and superior colliculus provides a constant, stimulus-independent factor pushing for each eligible action. The final decision is taken via softmax.
05612
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 07/06/2026
I’ve seen both sides: labs that make real efforts to share reusable code/data, and others where openness was discouraged by financial interests or competitive advantage. Industrial work is valuable and gives us many tools we use, but it shouldn’t come at the expense of scientific transparency.
020
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 07/06/2026
This reflects the importance of transparency IMHO. If data, code, parameters, preprocessing choices, etc. are shared, researchers can understand where differences come from, test robustness, and gradually converge toward better practices. Without that, disagreements stay hidden inside labs.
131
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 07/06/2026
Interesting article on something I think is very relevant to open science: several teams analyzed the same ephys dataset and got quite different answers. Not because the analyses were bad, but because this kind of work involves many analysis choices, and those choices can change the result.
11910
Reposted by Samuel Le Meur-Diebolt
Antoine Bergel @antbergel.bsky.social · 24/04/2026
⚠️ One week left to apply to our postdoc opening at the Paris Brain Institute ! If you have a background in neuroscience 🧠 , looking for a vibrant interdisciplinary environment in the heart of Paris, come join us !! All details below ⬇️ ⬇️.
032
Reposted by Samuel Le Meur-Diebolt
Agnès Landemard @agnesland.bsky.social · 15/04/2026
How does blood flow relate to brain activity? We discovered that it reflects two neural populations affected oppositely by arousal. Together, they explain neurovascular coupling in all brain regions and brain states! Out today in Nature: rdcu.be/fdC2A @uclbrainscience.bsky.social
The supply of blood to brain tissue is thought to depend on the overall neural activity in that tissue, and this dependence is thought to differ across brain regions and across brain states. However, studies supporting these views have measured neural activity as a bulk quantity and related it to blood supply following disparate events in different regions. Here we measure fluctuations in neuronal activity and blood volume across the mouse brain, and find that their relationship is consistent across brain states and brain regions but differs in two opposing brainwide neural populations. Functional ultrasound imaging (fUSI) revealed that whisking, a marker of arousal, is associated with brainwide fluctuations in blood volume. Simultaneous fUSI and Neuropixels recordings showed that neurons that increase activity with whisking have distinct haemodynamic response functions compared with those that decrease activity. Their summed contributions predicted blood volume across states.Brainwide Neuropixels recordings revealed that these opposing populations coexist in the entire brain. Their differing contributions to blood volume largely explain the apparent differences in blood volume fluctuations across regions. The mouse brain thus contains two neural populations with opposite relations to brain state and distinct relationships to blood supply, which together account for brainwide fluctuations in blood volume.
314462
Reposted by Samuel Le Meur-Diebolt
Julie Fabre @juliefabre.bsky.social · 12/04/2026
Want to explore connectivity & projection patterns yourself, like we do here? We released brain_street_view to let you pick any injection site in the Allen Connectivity Atlas and visualize where it projects in your favorite region of interest: github.com/Julie-Fabre/brain_street_view
03211
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
P.S. There is also a growing body of work suggesting that global brain activity can accompany arousal and motion. Steinmetz et al., 2019 (doi.org/10.1038/s415...) is a great example. And this very recent preprint from @mace-lab.bsky.social may shed more light on this! bsky.app/profile/mace...
doi.org
Distributed coding of choice, action and engagement across the mouse brain - Nature
Recordings from 30,000 neurons in 42 brain regions are used to delineate the spatial distribution of neuronal activity underlying vision, choice, action and behavioural engagement in mice.
030
Reposted by Samuel Le Meur-Diebolt
Emilie Macé @mace-lab.bsky.social · 16/03/2026
1/8. New preprint! Using fUSi in head-fixed mice🐭, we found that arousal events trigger a brain-wide wave of activity 🌊🧠. Surprisingly, this pattern was preserved during opto manipulations of the locus coeruleus, pointing to a minor role for noradrenergic tone. www.biorxiv.org/content/10.6...
23412
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
This study builds on a lot of important work from the fMRI field on understanding and mitigating motion artifacts. Huge thanks to all my co-authors, and to @iconeus.bsky.social for funding my PhD. This was very much a team effort! 🚀 12/12
120
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
That is why we see this paper as a framework rather than a final answer. We hope it helps make awake fUSI analyses more robust, while also motivating more validation and refinement in future work. 11/12
110
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
However, a key limitation is that movement can affect the signal through multiple ways. Some effects are likely direct artifact, but others may reflect genuine movement-related brain or vascular responses. Distinguishing those more cleanly will need additional ground truth in future work! 10/12
110
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
From there, the final step is to turn the benchmark into practical recommendations. We define optimized denoising strategies that perform more consistently across datasets than the standard workflow, giving a more robust starting point for awake fUSI analyses. 9/12
Optimized denoising paradigms preserve bilateral connectivity patterns and reduce artifactual anti-correlations in 2D seed-based maps compared to the standard strategy. Group-level seed-based maps are shown for the primary somatosensory cortex, hippocampus, and thalamus. Group-level maps were computed by averaging correlations (with Fisher transformation) within each mouse. Transparency thresholding was then applied, keeping significant values opaque while adjusting the opacity of non-significant values relatively to their 
-value (⁠
-test on Fisher-transformed correlations, family-wise error rate (FWER) controlled at 5%). Difference maps between high- and low-motion groups are computed by subtracting the average seed-based maps of the high-motion group from the low-motion group.
110
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
When we combine both metrics, only a small fraction of preprocessing strategies perform well overall. That is the main practical message: preprocessing choices have a major impact on the robustness and interpretability of awake fUSI results. 8/12
Only 26 of 792 denoising strategies achieve both satisfactory motion robustness and FC pattern preservation, all employing adaptive SVD clutter filtering. Metric values from Figures 5 and 6 are summarized into a single scatter plot. The same plot is shown four times, with color corresponding to the different denoising steps. The inset axes show strategies that pass the thresholds for both metrics. The four global strategies and the standard strategy are marked by stars.
110
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
We tested a large family of denoising strategies by varying clutter filtering, scrubbing, frequency filtering, and confound regression, then evaluate each strategy with two complementary metrics: 1. How robust it is to motion. 2. How well it preserves functional connectivity structure. 7/12
Outline of the denoising strategies and benchmark metrics. (a) General outline of the data denoising strategies benchmarked in this study. (b) Following the general outline in (a), each set of parameters was used to denoise post-QC acquisitions from all datasets (Table 1). Each denoised acquisition was then used to compute 6 seed-based maps. (c) The motion robustness metric was computed for each set of clutter filter, frequency filter, and confound regression parameters. The motion robustness metric is the average MSE across mice between seed-based maps computed with vs. without high-axial-velocity volumes. (d) The FC similarity metric was computed for each strategy and each dataset. The functional similarity metric is the average Dice similarity coefficient across mice between subject-level seed-based maps and the group-level seed-based map computed from the standard denoising strategy.
110
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
Once the problem is characterized, the key question becomes practical: which preprocessing choices best reduce motion-related bias without damaging functional connectivity? That is what we benchmark next. 6/12
110
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
With that initial motion characterization, the next question is: how do we detect artifacted acquisitions in practice? To answer this, we propose QC metrics that capture different signatures of motion contamination and help identify acquisitions that could lead to distorted connectivity. 5/12
Quality control metrics reveal distinct signatures of motion artifacts. Spatial and temporal QC metrics are displayed for two example acquisitions from dataset 1 with low (a) and high (b) motion artifacts. Both acquisitions were computed using the static SVD60 clutter filter. (Top) The top row shows the average power Doppler, CV, and brain axial velocity images. Both acquisitions show loss of power Doppler intensity in lateral and medial parts of the brain, typical of transcranial acquisitions. Both acquisitions also show high CV and high average axial velocity under the brain, where strong flashing artifacts can be observed during movement in power Doppler films. The high motion artifacts acquisition shows higher CV on lateral parts and on the brain edge, and higher axial velocity inside the brain. (Middle) The middle row shows the rGS, GV, and rDVARS signal. Note the co-occurrence of bursts in all three metrics, indicating that bursts of rGS (and thus rDVARS) are associated with high axial velocities. The high motion artifact acquisition shows more of the bursts, with higher amplitude in average. (Bottom) The bottom row shows three example seed-based maps with seeds placed in the primary somatosensory cortex, hippocampus, and thalamus. Seed-based maps were computed using the standard denoising strategy (static SVD60 clutter filtering without brain mask, scrubbing using rGS above 10%, [0.01, 0.1] Hz band-pass filtering, and GSR). Seed-based maps are thresholded to keep 10% of voxels with highest correlation values. Maps are overlaid on the Allen Mouse Brain CCFv3 template. The high motion artifacts acquisition shows loss of bilateral connectivity with more anti-correlated artifacts.
110
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
One nuance: the delayed post-motion signal may not just be a "delayed artifact". Its ~1.6 s peak is in the range of the awake-mouse fUSI HRF (e.g., see Nunez-Elizalde et al., 2022), raising the possibility that part of it reflects genuine neuronal activity. 4/12
110
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
Moreover, motion does not just create an immediate artifact. Using a FIR model, we show that brain motion has both immediate and delayed effects on power Doppler signals. That delayed effect is important: it means the relationship between motion and fUSI is not just a single-frame artifact. 3/12
Brain movement induces both immediate and delayed effects on power Doppler signals. (a) An FIR model was used to estimate the effect of the brain velocity on power Doppler intensity. Subject-level models were fitted for each mouse of dataset 1, and a group-level model was then fitted across mice. Resulting maps are shown for representative lags (see Supplementary Fig. S3 for all lags). At lag 0, brain movement leads to large increases in power Doppler signal across most of the FOV, as seen in Figure 1. However, significant effects are also observed in directly preceding and following frames. A delayed effect is then observed with increases in power Doppler signal across most of the vasculature, peaking at 1.6 seconds in the cortex and hippocampus and lasting up to 12 seconds in the thalamus. (b) Average 
-score time series are shown for three ROIs in the thalamus, hippocampus, and primary somatosensory cortex, displayed as colored circles in a. Shaded areas represent the 95% confidence interval of the mean.
130
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
The first point of the paper is that motion artifacts have a fairly specific signature. They show up with distinct spatial patterns and drive large, synchronous surges in power Doppler signal across most of the field of view. 2/12
Motion artifacts show distinct spatial patterns with large and synchronous power Doppler surges across the whole FOV. (a) Average power Doppler and axial velocity frames across dataset 1 during high vs. low power Doppler intensity periods. Low- and high-intensity frames were selected as the 5% lowest, respectively highest, average power Doppler intensity frames. High-intensity power Doppler frames show large surges in inferior cerebral veins, around the skull, and in lateral areas closer to muscles. In contrast, axial velocity frames show increased velocities in ventral brain regions and areas below the brain, while the cortex and dorsal skull areas show lower velocities. The lower part of the skull is particularly visible in the relative increase. (b) Carpet plot of standardized power Doppler time series from an example acquisition of dataset 1 (top) and corresponding GV (bottom). Large artifacts appear simultaneously across all brain voxels during high velocity periods.
120
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 10/04/2026
New paper out 🎉 Awake fUSI is powerful, but motion can strongly bias the data, even in head-fixed experiments. In this paper, we tried to systematically characterize those artifacts, benchmark denoising strategies, and turn that into practical recommendations for awake fUSI of mouse brains. 1/12
1166
Reposted by Samuel Le Meur-Diebolt
Antoine Bergel @antbergel.bsky.social · 10/04/2026
⚠️ Postdoc Opportunity in System Neuroscience at the Paris Brain Institute ⚠️ Feel passionate about neuroscience and sleep 🧠? Want to learn cutting-edge techniques 🔬 to answer exciting basic science questions? Want to work in a multidisciplinary environment at the heart of Paris 🏨? Come join us !
11613
Reposted by Samuel Le Meur-Diebolt
SWC/GCNU Neuroinformatics Unit @neuroinformatics.dev · 13/03/2026
We have three neuroscience research software engineering jobs available at @sainsburywellcome.bsky.social, two joint with the Advanced Microscopy Facility. All positions start ASAP, funded until Sept 2028 in the first instance. Salary £54-64k. More details: neuroinformatics.dev/get-involved
neuroinformatics.dev
Get involved — NIU
11111
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 26/02/2026
Next #fUSI - #BIDS meeting is coming up! 🧠 📅 March 19, 5PM CET 🎯 Solving remaining spec remarks & moving toward official maintenance. Newcomers welcome! 🚀 Planning to BIDSify your data? Join us, we’re here to answer your questions. Details on the GDoc: docs.google.com/document/d/1...
docs.google.com
fUSI-BIDS extension proposal v0.0.12
BIDS Extension Proposal 40 (BEP040): fUSI-BIDS version 0.0.12.dev Available under the CC-BY 4.0 International license. Extension moderators/leads: Jean-Charles Mariani <jean-charles.mariani@cri-paris...
042
Reposted by Samuel Le Meur-Diebolt
Institute Physics for Medicine Paris @physmedparis.bsky.social · 13/01/2026
𝗞𝗶𝗰𝗸𝗶𝗻𝗴 𝗼𝗳𝗳 𝟮𝟬𝟮𝟲 𝘄𝗶𝘁𝗵 𝗮 𝗻𝗲𝘄 𝗽𝘂𝗯𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗶𝗻 𝗖𝗲𝗹𝗹 𝗥𝗲𝗽𝗼𝗿𝘁𝘀 @felipecybis.bsky.social, postdoct in our lab, is the first author of the study "A vascular code for speed in the spatial navigation system", now published in Cell Reports: doi.org/10.1016/j.celrep.2025.116791 #Ultrasound #fUS #Research
076
Samuel Le Meur-Diebolt @sdiebolt.bsky.social · 13/11/2025
Long-awaited #fUSI - #BIDS meeting scheduled! 🧠 📅 Nov 27, 5PM Paris time 🎯 Finalizing BEP for BIDS maintainer discussions Implemented fUSI-BIDS or planning to? We need your input to strengthen the spec. Join us! Details on the specification GDoc (docs.google.com/document/d/1...)
docs.google.com
fUSI-BIDS extension proposal v0.0.11
BIDS Extension Proposal 40 (BEP040): fUSI-BIDS version 0.0.11 Available under the CC-BY 4.0 International license. Extension moderators/leads: Jean-Charles Mariani <jean-charles.mariani@cri-paris.org...
021
Reposted by Samuel Le Meur-Diebolt
Emilie Macé @mace-lab.bsky.social · 11/09/2025
Thrilled to share that our work is now published in Science! ✨ We found a preference for visual objects in the mouse spatial navigation system where they dynamically refine head-direction coding. In short, objects boost our inner compass! 🧭 www.science.org/doi/10.1126/... 🧵1/
817572