iiinova — where your digital self keeps evolving

Léo Varnet @leovarnet.bsky.social · 09/09/2026

Congratulations to my students Azal Le Bagousse and Mauro Manucci for their outstanding presentations at Forum Acusticum #FA2026 this morning, despite the very early schedule. If you want to read more about their work: hal.science/hal-05693617 hal.science/hal-05679708

030

Léo Varnet @leovarnet.bsky.social · 01/08/2026

"Trois ans de recherches de financements", quelques conclusions tirées de mes échecs aux différents appels à projets. dbao.leo-varnet.fr/2026/08/01/t...

The House of Dust, Alison Knowles. Une imprimante qui génère des pages et des pages de poésie automatique.

121

Léo Varnet @leovarnet.bsky.social · 21/07/2026

We tested two variants of the method, using 3 variable parameters or 6. Both outperformed reverse correlation: 600 trials (≈40 min) achieved comparable precision to hours of traditional testing. Next steps: running simulations and collecting new data to prepare a journal submission.

100

Léo Varnet @leovarnet.bsky.social · 21/07/2026

This adaptive method efficiently probes the mental distribution of vowels: after convergence, we can map the distributions along any dimension. The plot below shows clear, vowel-specific clusters in the F1-F2 space. Different vowels correspond to different regions of the mental space. (5/6)

Figure 5 from the paper: Heatmaps of the counts of selected stimuli pooled across participants in F1-F2 space. Each panel corresponds to a target vowel: "a" (left), "e" (center), "u" (right). We clearly see that the selected stimuli are gathered in different regions for the 3 target vowels.

100

Léo Varnet @leovarnet.bsky.social · 21/07/2026

This results in chains of selected sounds that converge toward a prototypical representation of the target vowel in the acoustic space. (4/6)

Figure 2 from the paper. Trajectory example for two chains from a single participant, in F1-F2 space. Both chains start at a random position in the upper part of the space, then converge to the lower part.

110

Léo Varnet @leovarnet.bsky.social · 21/07/2026

The selected sound becomes the new reference, and we propose a new variant nearby in parameter space... and we iterate the process like a Metropolis-Hastings algorithm. (3/6)

100

Léo Varnet @leovarnet.bsky.social · 21/07/2026

We need a smarter approach to explore the high-dimensional perceptual space of vowels using a perceptual experiment. (2/6)

100

Léo Varnet @leovarnet.bsky.social · 21/07/2026

Vowels are complex: they are defined by multiple acoustic parameters (formants F1-F4, pitch, duration, etc.).

The IPA vowel trapezium, with the relative positions of most vowels

101

Léo Varnet @leovarnet.bsky.social · 21/07/2026

We explored how listeners mentally represent vowels, i.e. which acoustic properties are most critical for identification. hal.science/hal-05679708v1/document (1/6)

100

Léo Varnet @leovarnet.bsky.social · 21/07/2026

Here's a thread about a nice student-led project on vowel perception that evolved into a full conference paper for Forum Acusticum 2026 in Graz. #FA2026

First page of the paper with title authors and abstract.

152

Léo Varnet @leovarnet.bsky.social · 03/07/2026

1177/23312165261465060 Huge thanks to Lily Paulick for transforming a very complex set of data into a clear and rigorous paper! #reversecorrelation #audition #hearing #psychoacoustics

010

Léo Varnet @leovarnet.bsky.social · 03/07/2026

Can you discriminate a rising frequency sweep from a downward one, in the presence of a background noise? Using reverse correlation, we investigated listeners’ sensitivity to specific frequency sweeps and the auditory mechanisms involved. journals.sagepub.com/doi/full/10.

120

Léo Varnet @leovarnet.bsky.social · 30/04/2026

Bonus: I updated the French hybrid nouns categories on the French Wiktionary to include all the crowd-sourced suggestions: fr.wiktionary.org/wiki/Cat%C3%... fr.wiktionary.org/wiki/Cat%C3%... (10/10)

fr.wiktionary.org

Catégorie:Mots masculins en français pouvant désigner des femmes — Wiktionnaire, le dictionnaire libre

000

Léo Varnet @leovarnet.bsky.social · 30/04/2026

Moreover, there are other arguments in favor or against gender-fair writing. But this study adds another piece to the puzzle of how grammatical gender shapes mental representations, even when this information is in fact totally irrelevant as in the case of hybrid nouns. (9/10)

100

Léo Varnet @leovarnet.bsky.social · 30/04/2026

It takes the brain an average of 130 ms to update its default masculine representation and realize that "un être humain" can in fact be a woman. Note that this does not mean you should stop using masculin hybrid nouns altogether. They are still less biased than generic masculine forms. (8/10)

100

Léo Varnet @leovarnet.bsky.social · 30/04/2026

Surprisingly, the results indicate that masculine hybrid nouns (such as "un être humain", "un nouveau-né", "un ange") introduced a masculine bias during reading, while feminine hybrid nouns (such as "une personne", "une vedette", "une recrue") did not! (7/10)

Figure 1 from the paper: Average reaction times and percentage of correct responses for each participant in each of the two conditions.

100

Léo Varnet @leovarnet.bsky.social · 30/04/2026

These terms are useful when you don't want to specify the gender of the person you're referring to. We started by crowd-sourcing a list of French hybrid nouns via a collaborative notepad on Mastodon. Huge thanks to everyone who contributed! (6/10)

110

Léo Varnet @leovarnet.bsky.social · 30/04/2026

Here, we decided to apply the same methodology to another form of gender-fair writing: hybrid nouns. Hybrid nouns are words with a fixed grammatical gender which can refer to either a man or a woman (e.g. "un individu", "une personne", etc.). (5/10)

100

Léo Varnet @leovarnet.bsky.social · 30/04/2026

We found that the middle dot forms eliminated masculine bias, but gender-unmarked forms still elicited slightly biased representations. See more about this study here: fediscience.org/@LeoVarnet/1... (4/10)

fediscience.org

Léo Varnet (@LeoVarnet@fediscience.org)

Attached: 1 image Our latest article on #GenderFairLanguage has been published in Frontiers in #Psychology! You can read it #OpenAccess here: https://www.frontiersin.org/articles/10.3389/fpsyg.2023.1...

100

Léo Varnet @leovarnet.bsky.social · 30/04/2026

One possible solution is gender-fair writing. In a previous study, we tested whether gender-unmarked forms (such as "l'athlète", which can be masculine or feminine) or middle dot forms efficiently reduce the masculine bias observed in generic masculine forms. (3/10)

100

Léo Varnet @leovarnet.bsky.social · 30/04/2026

Quick recap. Generic masculine forms are theoretically neutral (in French you can use "les musiciens" to refer to male and female musicians), but empirical studies show that when we read "les musiciens", our brains default to a masculine interpretation, and only later adjust to include women. (2/10)

100

Léo Varnet @leovarnet.bsky.social · 30/04/2026

Our follow-up study on the psycholinguistics of gender-fair language is finally out! This was a long journey, so here's a thread to explain what we did and what we discovered. The #openaccess article is here for those who want the full story: hal.science/hal-05607624 (1/10)

Title page of the article "Grammatical Gender Biases Interpretation of Masculine, but Not Feminine, Hybrid Nouns"

152

Léo Varnet @leovarnet.bsky.social · 23/04/2026

I uploaded the preprint on #HAL and submitted it to @psych.peercommunityin.org for a journal-independent peer-review. Very excited to see how it goes! (here's a link to the preprint if you want to have a look: hal.science/hal-05594953v1) (2/2)

hal.science

Making sure you're not a bot!

000

Léo Varnet @leovarnet.bsky.social · 23/04/2026

I wrote my first single-author manuscript, reviewing evidence for an overlooked effect of noise on speech perception. Since no junior researcher were involved (so no pressure to secure traditional publications for their CVs), I decided to explore alternatives to the standard publishing system. (1/2)

192

Reposted by Léo Varnet

Julia M. Rohrer @dingdingpeng.the100.ci · 21/03/2026

Psychology has a whole cottage industry in which people come up with some construct that is essentially "attitudes/beliefs/expectations/feelings about X", and then the central claim is that this construct is a super important determinant of future X outcomes.>

1416837

Reposted by Léo Varnet

Kevin Cariou @cariouk.bsky.social · 13/03/2026

Les deux collègues chimistes assassinés par Israël ont été délibérément ciblés par l'armée sous prétexte qu'ils auraient été membres du Hezbollah. Condamnés et exécutés en toute impunité. Feu Hussein Bazzi était un alumnus de @sorbonne-universite.fr où il avait effectué sa thèse.

franceinfo.fr

Guerre au Moyen-Orient : au Liban, deux enseignants tués dans une frappe israélienne sur un campus de l'université publique

Israël affirme que le métier des deux victimes était une couverture et qu'ils étaient liés au Hezbollah. Le recteur de l'université dénonce un montage photo publié par un site israélien, qui est censé...

174

Léo Varnet @leovarnet.bsky.social · 26/02/2026

Suite du chapitre 4, on rentre dans le vif du sujet : l'exploration de spectrogrammes de sons de parole permet de se rendre compte que la perception des phonèmes n'est pas un mécanisme aussi simple qu'il n'en a l'air au premier abord. leovarnet.github.io/psychoac-man...

010

Reposted by Léo Varnet

Léo Varnet @leovarnet.bsky.social · 16/01/2026

To wrap up #SPIN2026 in a playful way, I prepared a psycholinguistics-related blind test for the last evening event. The goal was to identify both the song and type of signal processing that was applied to it. Let's see if you can solve these ⬇️

111

Léo Varnet @leovarnet.bsky.social · 16/01/2026

The Speech in Noise (SpiN) workshop, a two-day scientific meeting which brings together researchers from several European hearing-research groups working on speech-related topics. 2026.speech-in-noise.eu?p=informatio...

2026.speech-in-noise.eu

SpiN2026 - 17th Speech in Noise Workshop - Paris, France

The SpiN workshop is organized every year by a European research group within the speech and hearing community, and consists of two days of talks by invited speakers (PhD students, postdocs, and princ...

120

Léo Varnet @leovarnet.bsky.social · 16/01/2026

111

Léo Varnet @leovarnet.bsky.social · 15/01/2026

That's it, #SPIN2026 is over, thank you to all attendees for two great days of thought-provoking talks, amazing poster sessions... and actual speech-in-noise demo during coffee breaks!

Conference poster, titled "speech in noise workshop"

1516

Léo Varnet @leovarnet.bsky.social · 12/01/2026

What an incredible start for the #SpiN2026 workshop! Thank you SpiNners, day 1 was great, I'm looking forward to day2!

A crowded room full of posters and workshop attendees

011

Reposted by Léo Varnet

Sam Gershman @gershbrain.bsky.social · 09/01/2026

With some trepidation, I'm putting this out into the world: gershmanlab.com/textbook.html It's a textbook called Computational Foundations of Cognitive Neuroscience, which I wrote for my class. My hope is that this will be a living document, continuously improved as I get feedback.

16590237

Léo Varnet @leovarnet.bsky.social · 08/01/2026

The program booklet for the SpiN workshop is now ready to download! Two days of intense discussions on psychoacoustics and psycholinguistics await at ENS Paris next week! 2026.speech-in-noise.eu/files/SPIN2…

Cover of the booklet. Title: Speech in Noise Workshop, Abstracts. With a colorful picture of the Eiffel Tower.

052

Léo Varnet @leovarnet.bsky.social · 17/12/2025

I gave it my best shot, but the 'Longest Experiment of the Year' award from the @lsp-ens.bsky.social slipped through my fingers this year... Congratulations to the winners!

030

Léo Varnet @leovarnet.bsky.social · 02/12/2025

L'écriture de mon manuel de psychoacoustique #openaccess avance bien : chapitre 1 (méthodologie de la psychophysique) terminé, chapitre 2 (physiologie du système auditif) terminé pour ce qui concerne le système auditif périphérique, j'attaque le chapitre 4 ! leovarnet.github.io/psychoac-man...

leovarnet.github.io

Bienvenue ! — Psychoacoustique et perception auditive, une introduction

011

Léo Varnet @leovarnet.bsky.social · 28/11/2025

Many different experiments are included natively in the toolbox (in particular you can run the experiments from all our previous studies). We also conducted a replication of a 50-year-old experiment to demonstrate the versatility of the toolbox.

100

Léo Varnet @leovarnet.bsky.social · 28/11/2025

The toolbox is also designed to be as “plug-and-play” as possible, allowing researchers unfamiliar with the reverse-correlation approach to easily apply it to their own research questions. Finally, fastACI is open source, supporting transparent, replicable, and computationally reproducible research.

110

Léo Varnet @leovarnet.bsky.social · 28/11/2025

By integrating all these options within a consistent and flexible framework, the fastACI toolbox enables, for the first time, systematic comparisons across different configurations.

100