Pieter Barkema | Research Scientist & Engineer

Postdictive Perception in Primary Visual Cortex

Mon, 01 Dec 2025 00:00:00 +0000

The Science of Postdiction

Do brain representations reflect what was presented to our eyes (retinal input) or what we actually perceive? This project investigates Postdiction, a phenomenon where information arriving later influences how we perceive an earlier event.

Research Highlights & Technical Achievements

Advanced Neuroimaging: Used 7T fMRI to achieve layer-specific resolution in the primary visual cortex (V1).
Layer-Specific Decoding: Discovered that deep layers of V1 encode subjective perception (illusory flashes), while middle layers continue to track the objective retinal input.
Top-Down Modulation: Identified that postdictive perception is likely facilitated by feedback from superior temporal cortex to the early visual cortex.
Project Management: Led a £400K project end-to-end, managing a team of three students and coordinating data collection for over 300 participants.
High-Impact Communication: Presented findings at the Visual Science Society (VSS) 2025.

Scientific Impact

Our findings complete a feedback model of sensory processing across seeing and hearing, suggesting that early visual cortex is not just a passive relay station but an active participant in constructing hallucinations and perception.

Family Talk Royal Institution

Sat, 13 Sep 2025 13:00:00 +0000

Layer-specific fMRI of Postdictive Perception at 7T

Thu, 15 May 2025 10:30:00 +0000

Cross-Category Information (CCI)

Sat, 01 Jun 2024 00:00:00 +0000

The Challenge: Signal vs. Noise

In neuroimaging, “noise” is often defined as trial-to-trial variability that doesn’t match the stimulus. However, much of this variability is shared across neuronal populations and may represent top-down feedback or internal predictions.

The Innovation: The CCI Metric

I invented the Cross-Condition Information (CCI) metric to quantify how much of this “noise” is actually structured and informative. With Manifold Learning (PCA) and Subspace Alignment, CCI measures the overlap between variation in neural representation spaces across objects from the same category (i.e. different animals).

Technical Implementation

Manifold Alignment: Developed algorithms to align low-dimensional latent representations.
Noise Decoding: Demonstrated that trial-by-trial variability is not stochastic but follows a specific geometric structure that predicts behavioral outcomes.
High-Performance Computing: Optimized the pipeline in Python (NumPy/SciPy) to process multi-terabyte datasets across HPC clusters.

Scientific Impact

CCI provides a tool for researchers to interrogate how informative noise in the brain is for object classification in human vision.

PCNportal: Scalable Normative Modelling

Tue, 01 Aug 2023 00:00:00 +0000

Built with Flask, Docker, and Gunicorn, this platform allows clinical researchers to apply hierarchical Bayesian models to individual brain data at scale.

Mon, 01 Jan 0001 00:00:00 +0000