Manuscript submitted: Leo di caprio looking young and healthy in Titanic Manuscript accepted: Leo di caprio looking like he's at death's door in The Revenant
Just had a paper accepted after 4 rounds of revisions and *10* reviewers!
"Preprint servers are a time machine, they move everyone forward 12 months and speed up the exchange of ideas"
ht @pedrobeltrao.bsky.social www.evocellnet.com/2021/06/a-no...
"Frame Effects Across Space and Time" is published: doi.org/10.1167/jov....
The effect:
- extends a bit over space,
- not time,
- mostly depends on frame edge positions
- doesn't decrease with experience
Looks like vision uses references to localize objects in space.
Personally, I try to come up with sensible hard cutoffs. E.g. if a directions is more than 90 degrees of, you're not even close to going the right way, or if your RT is 20 seconds, you were probably doing something else anyway.
Yeah, I'm on the fence about that. The output here has aggregate data that can be used to do an ANOVA (and family) in your favourite stats software. So there is no way to remove outliers afterward from that file. But... no one has to use the feature, and you can always go back and run it again.
I'll be looking for bugs and may update this periodically. If you see any, please let me know.
mthart.shinyapps.io/PreProcessor/
So I hesitate to put this out here, but on the other hand, I should probably show graduate students how to use their understanding of code to make things like this and be competitive in the current job market.
Vibe coded a Shiny app for data preprocessing. Meant as a tool for undergrads doing projects in the lab. Computer literacy is low across the board, and with tools like these (and the LLMs to build them) there's no need to understand what you're doing.
8/8 Finally published! 🚀
This project started with pilot data in 2016 and came out in 2026 — nearly a decade in the making. Proof that good data doesn’t expire. If your findings still matter: publish them.
This graph shows 4 scatter plots. At the top it shows how well active as well as passive localization can predict implicit reach aftereffects, and at the bottom it shows how well a combination of variance and shift of localization can predict reach aftereffects for both active and passive reach aftereffects. The R^2 for using localization shift are 0.24 and 0.20, and when adding the standard deviation of baseline localization this increases to 0.28 and 0.29 for active and passive localization respectively.
7/8 Proprioceptive recalibration can't be predicted either. But reach aftereffects can be! Both by precision of hand localization and by the level of proprioceptive recalibration. These predictors are independent, hinting at two ways proprioception guides implicit adaptation.
6/8 In a subset of participants we also check whether or not any measures of motor learning can be predicted by any of the measures of precision: testing the exploration-exploitation hypothesis. We see no way to predict either rate of change or asymptotic levels of reach adaptation.
This graph show the standard deviations of individual participants in both our younger and older participants on training reach direction, no-cursor reach directions and active as well as passive localization, with a mean and 95% CI as well as a distribution. The distributions for the two age groups look very similar.
5/8 We also look at whether or not age decreases the precision of reaches or hand localization, and find that is not the case. The effects of age on motor performance may just not be as dramatic as often reported.
Scatter plot showing SD of passive (proprioception only) over active localization (proprioception + prediction) of 270 participants, with an orthogonal distance regression (slope = 1.09; R^2 = 0.67). There are slightly fewer dots above the unity line than below.
4/8 What we can say now is that predictive (efferent) signals do contribute to hand localization in this task, but the contribution is really small (~8%, slightly larger SD in passive localization).
3/8 In our previous papers, it seemed like we didn't have enough data to look at this.
From 2016: dx.doi.org/10.1371/jour...
From 2019: doi.org/10.1371/jour...
(We're still using the same Poser images as in 2016…)
2/8 Our original question was inspired by other work where we tried separating predictive and proprioceptive signals contributing to hand localization. When participants move their own hand they have both signals, but when the robot moves their hand they only have proprioception.
1/8 Paper! Using previously published data from 270 participants, we look at precision of reaches and hand proprioception in rotation adaptation:
doi.org/10.1186/s130...
📆 updated for 2026!
list of summer schools & short courses in the realm of (computational) neuroscience or data analysis of EEG / MEG / LFP: 🔗 docs.google.com/spreadsheets...
(Messages above by Raphael, who is not on BlueSky, and worked on this project since 2019... congrats on - more or less - finishing it!)
Overall, we show that preparatory and feedback-related EEG markers differ across perturbation types, showing distinct neural processes across motor learning types. Data is also available online, and we welcome feedback!
For feedback processing, P3 amplitude reduces across training for the rotation but remains stable for the mirror reversal. We also observe a sustained positivity that seems to index implicit learning.
We identify EEG markers distinguishing these two motor learning types across temporal and frequency domains. For movement preparation, the Readiness Potential shows training-related modulation for the rotation, but not the mirror reversal. Beta and alpha modulation was also observed in the rotation.
New preprint out! (doi.org/10.64898/202...)
We examined how EEG activity during movement preparation and error processing distinguishes learning and error processing between motor adaptation and de novo learning.
EEG Dynamics of Movement Preparation and Error Processing Distinguish Motor Adaptation From De Novo Learning https://www.biorxiv.org/content/10.64898/2025.12.11.693740v1
Someone in my lab asked me to make a QR code, yet again. So I vibe-coded them a Shiny app, and now you could use it too: mthart.shinyapps.io/ShinyQR/
Vibe coding is fun, but I would not use it for data analysis.
Source code: github.com/thartbm/Shin...
I think almost all scientific projects should be planned carefully. And I think an app can dramatically improve that. So I wrote an app for that (free for now, if you can fund this let me know). I tested it quite a bit (>8000 users in beta so far). try it: planyourscience.com
For sure! We're developing a task where we can manipulate the environmental aspect much better. Will be a while, but it's coming.
This data was recorded in 2022 or so... I finally got my act together to finish all the data analyses. But... data does not expire!
If you want to get your 2 cents in before we get reviews from the journal, now's your chance!
We find that the influence of the frame extends somewhat in space, but not in time. Motion signals don't play much of a role; the frame effect is dominated by the perceived position of the frame. This tells us that our visual system localizes objects relative to their reference frame.
Here we test what happens when there are offsets in space and time between the frame and probes. And technically also effects of two motion signals apart from those of the frame. Videos linked in the pdf!
