Cognitive neuroscientist · Duke University
How does the brain reorganize mental representations when familiar ways of thinking no longer work—and how does this change support future learning?
My research on flexible cognition examines how representational change enhances memory, how knowledge supports discovery, and how interventions change cognitive performance. Within this work, my theoretical framework proposes that how rapidly prediction error is reduced shapes the subjective experience of discovering a solution. Current follow-up fMRI findings link stronger Aha! experiences to steeper cortical representational updating.
Methods: fMRI, representational similarity analysis (RSA), behavioral experiments, eye tracking, and computational approaches.
Using Mooney images and RSA, I quantified representational change in visual cortex before and after a solution. Stronger insights involved greater neural reorganization—even though the image stayed unchanged. This cortical change and hippocampal activity predicted later memory.
How do existing (neural) representations shape the discovery of new connections? Current findings suggest that stronger semantic organization in object-selective visual cortex predicts later solution discovery—even before the problem is posed.
How can interventions change cognition? I study how training and pharmacological interventions influence attention, memory, and brain connectivity. In a randomized study, video-game training improved sustained attention in both healthy adults and patients with schizophrenia.
At Duke University, I work with Roberto Cabeza as a Research Scholar. My background spans psychology, cognitive neuroscience, and the history and philosophy of science. I received my PhD in Cognitive Neuroscience from the University Medical Center Hamburg-Eppendorf in 2020.
This interdisciplinary perspective shapes my interest in how new understanding emerges. From Kuhn’s scientific revolutions and Kekulé’s benzene ring to everyday insights, representing a problem differently can reveal a solution that previously escaped us. I want to understand how the brain makes that shift—and turns it into lasting learning.
As AI becomes better at generating explanations, understanding our own ability to question assumptions, discover alternatives, and judge new ideas matters more than ever.
My ORCID profileLook for an object, connect two images, rethink a word, or tackle a spatial puzzle. These short examples let you explore different routes to a new interpretation. Take your time before opening a hint or solution.
There is a real object in this black-and-white image. Can you find it?
The image stays the same, but your interpretation changes. In my fMRI study, I used RSA to measure how activity patterns in visual cortex reorganize before and after recognition—and how this relates to insight and later memory.
Read the neural representational-change study
Task materials & code
The dark shape near the centre is its nose. Look again: does the rest of the face fall into place?
The pixels have not changed. What may have changed is the interpretation that organizes them.
Original stimulus from Becker, Sommer & Cabeza (2025).
Find a third object that connects both pictures. Its appearance should resemble one cue; its meaning or function should relate to the other.
I developed the Language-Independent Remote Associate Task (LI-RAT) with Roberto Cabeza to study creative problem solving using pictures.
Read the task-development paper
Stimuli, task scripts & norms
Look at the shape of the corset and the function of the stopwatch.
It has the narrowed shape of a corset and measures time, like a stopwatch.
The corset becomes useful through its shape, rather than through its familiar function as clothing.
Example from Becker & Cabeza’s LI-RAT. Original task images.
First, picture a backdrop: the scenery behind a theatre stage. Keep that meaning in mind for a moment. Then open the word problem.
In my modified compound remote associate (CRA) tasks, a related word biases the meaning of one cue. This helps me separate restructuring from the subjective Aha! experience.
Read the modified-CRA paper
Open manuscript & paradigm description
Find one word that can go before each of these three words to form a familiar compound.
Does “drop” have to keep its theatrical meaning?
Raindrop. Raincoat. Rainproof.
“Drop” now refers to water, rather than stage scenery. Compare starting with dewdrop: it already suggests a meaning that fits the solution.
A changed meaning and an Aha! feeling need not always occur together. That distinction is central to my modified CRA research.
English demonstration using the example in Becker, Kühn & Sommer (2020); simplified from the experimental procedure.
Connect all nine dots with four connected straight line segments, without lifting your pen or retracing a line.
Choose a starting point on the board, then choose four endpoints. Each click or tap adds a straight segment.
Which limits are stated in the instructions, and which ones have you added yourself?
Focus the board. Move with the arrow keys; press Enter or Space to place a point. Backspace undoes a point and Escape resets the board.
A classic insight problem, included to explore the role of self-imposed constraints.
Choose a starting point.
Start at the top-left dot, go right beyond the top row, diagonally down-left, straight up, then diagonally down-right.
The instructions never confined your lines to the square suggested by the dots. Relaxing that constraint opens a route—but still leaves a problem to solve.
Classic nine-dot problem; this website’s drawing interface is an adaptation.
For research collaborations, academic enquiries,
or conversations about flexible cognition.
Center for Cognitive Neuroscience
Duke University
Durham, North Carolina, USA