CAOS 2024 - 5 | Rovereto, May 9-11 | Floris de Lange

CAOS 2024 - 5 | Rovereto, May 9-11 | Floris de Lange

by Floris de Lange

Life & Natural Systems Neuroscience PSANNeurosciencesPSAN5Cognitive and behavioural neuroscience
🎙 Floris de Lange 👥 2K 📅 November 13, 2025 ⏱ 70 min 👁 55 📄 expert opinion 🧭 2026-08-16
Available in: English (current) Français

Keywords

predictionvisual cortexstatistical learningsuccessor representationprediction error

Summary

Floris de Lange presents his research on predictive processing in visual cognition. He begins by highlighting the ubiquity of structure in the world, from language to visual scenes and actions, and introduces statistical learning as the brain’s ability to extract these regularities. He then outlines three questions: how far ahead the visual system predicts, what is contained in a prediction, and how prediction operates in naturalistic settings. Using fMRI with high temporal resolution, his lab demonstrated that presenting the first element of a learned sequence triggers activity along the expected trajectory, akin to ‘pre-play’ observed in rat visual cortex. They further showed that this predictive trace is temporally discounted and specific to future states, not past ones. To probe the content of predictions, they used paradigms where stimuli predicted others, finding suppressed responses for predicted inputs across visual areas, and that surprise signals emerge first in anterior regions before propagating to early visual cortex. By examining prediction errors, they aim to infer the nature of internal models. The talk concludes with ongoing work on naturalistic prediction, emphasizing the importance of understanding prediction in real-world contexts.

185 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides valuable insights into predictive processing, synthesizing findings from multiple studies and presenting novel data. The argumentation is solid, building logically from foundational concepts to specific experiments. De Lange effectively uses analogies (e.g., tennis) and references to prior work (e.g., Saffran, Stenfeld) to support his claims. He also acknowledges limitations, such as the confounding of space and time in early experiments, and proposes more rigorous tests. The presentation of preliminary data is appropriately cautious.

Scientific Rigor, Source Quality, Title Accuracy

The talk demonstrates high scientific rigor, with clear methodology and careful interpretation. De Lange cites relevant literature, including studies by Saffran, Stenfeld, and others, and references his own published work. The sources are credible and appropriately used. The title accurately reflects the content, focusing on prediction in visual cognition. The talk is part of a scientific workshop, and the content aligns with the title.

155 words

Title / Content Match

The title accurately reflects the content: a talk on prediction in visual cognition within a workshop on concepts, actions, and objects.

Quality & Reliability

8/10

Talk by a leading researcher in predictive processing, presenting peer-reviewed studies and preliminary data. High scientific rigor, but some claims are based on unpublished or preliminary results.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Concurring Sources

  • Predictive Processing: A Canonical Computation — Review article supporting the predictive processing framework.

Dissenting Sources

Contribution & Novelties

The talk presents original research on predictive processing, particularly the demonstration of pre-play in human visual cortex and the temporally discounted nature of predictive traces. It also offers a novel approach to probing the content of predictions via prediction errors. The work contributes to understanding how the brain uses statistical regularities to anticipate future sensory input.

Pour aller plus loin :

  • Predictive Processing — Overview of the theoretical framework.
  • Statistical Learning — General concept of learning regularities.
  • Successor Representation — Computational model relevant to the talk.
  • Population Receptive Field Mapping — Technique used in fMRI studies.

96 words

Radar Profile

The radar profile shows high scores across all dimensions, indicating a well-rounded and rigorous scientific presentation. The talk excels in information quantity and quality, with a strong technical level and high reliability.

Reliability 8/10

💬 No comments were provided for analysis.