Contributed Talks: “Audition and Language” - CCN 2025

Contributed Talks: “Audition and Language” - CCN 2025

🎙 Cognitive Computational Neuroscience 👥 4K 📅 October 8, 2025 ⏱ 55 min 👁 111 📄 conference talks 🧭 2026-08-15
Available in: English (current) Français

Keywords

language processingauditory cortexfMRIEEGinvarianceentropyprediction

Summary

This video is a recording of a contributed talks session at the Cognitive Computational Neuroscience Conference 2025 in Amsterdam, focusing on audition and language. The session includes five talks. The first talk by Greta Tuckute presents an ultra-high field fMRI study identifying two main dimensions of sentence processing shared across individuals: processing difficulty and meaning abstractness. These dimensions form a spatial topography in frontal and temporal language areas. The second talk by Akhil Bandreddi uses intracranial EEG to investigate how the brain encodes invariant representations of spoken words and talker identity. The researchers develop a method to extract invariant neural dimensions and show that these dimensions are robust and can be applied to whispered speech. The third talk by Pierre Guilleminot introduces a framework for decomposing uncertainty in speech processing using Rényi entropy, which captures both dispersion and strength of probability distributions. The fourth talk by Lars Kopel examines pupil-linked arousal as a marker of belief updating in a dynamic auditory environment. The fifth talk by Mengting Xu explores how auditory stimuli modulate visual integration via alpha-band oscillations. The session is technical and aimed at a specialized audience, with each talk presenting original research and methodological innovations.

197 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talks provide valuable insights into the neural mechanisms of language and audition. Tuckute’s study is notable for its use of 7T fMRI and a large set of sentences to derive interpretable dimensions of neural responses. Bandreddi’s work offers a novel method for quantifying invariance in neural representations, with potential applications beyond speech. Guilleminot’s introduction of Rényi entropy to model uncertainty in speech processing is theoretically motivated and could refine predictive coding models. The arguments are supported by empirical data and cross-validation, though the presentations are brief and some details are omitted. The speakers acknowledge limitations and future directions, indicating a rigorous approach.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is high: the studies use appropriate neuroimaging and statistical methods, and the results are presented with caution. The sources are primarily the speakers’ own research, with some references to prior work (e.g., Norman-Haignere). The title accurately describes the content as a session of contributed talks. No external sources are cited in the description, so the analysis relies on the talks themselves. The adequacy between title and content is perfect.

190 words

Title / Content Match

The title accurately reflects the content: a session of contributed talks on audition and language at CCN 2025.

Quality & Reliability

8/10

The video presents peer-reviewed research from established labs, with methodological details and cross-validation. However, it is a conference session with limited context and no external verification.

Key Moments

Cited Sources

  • No external sources cited in the video description. — The video description only lists the talks and speakers.

Concurring Sources

Contribution & Novelties

The talks present original research with methodological innovations. Tuckute’s work provides a novel two-dimensional framework for understanding language representations in the brain. Bandreddi’s method for extracting invariant neural dimensions is a new approach to studying invariance in neural codes. Guilleminot’s use of Rényi entropy offers a more nuanced measure of uncertainty than Shannon entropy. These contributions advance the field of computational neuroscience.

Pour aller plus loin :

94 words

Radar Profile

The radar profile shows high scores across all dimensions, indicating a technically rigorous and informative session. The balance between quantity and quality of information is strong, with a slight emphasis on technical depth.

Reliability 8/10