I use neuroimaging techniques to gain a deeper understanding of the brain’s role in memory and social cognition.

Project 1: Does stimulus predictability affect fMRI repetition suppression?

I analyzed fMRI data, collected by Per Sederberg and colleagues, to examine the impact of upcoming stimulus predictability on repetition suppression. Repetition suppression is a fMRI phenomenon where the neural response to a repeated stimulus is weaker than the response to a novel stimulus. While we observed the expected repetition effect, we did not find a detectable effect of the experimental manipulation of stimulus predictability.

Project 2: Can we use machine learning to predict memory scores from brain data in a large sample?

This study is also listed on my machine learning projects page.

I used resting state fMRI data from 540 individuals in the Human Connectome Project to investigate whether machine learning could be used to predict individual differences in recognition memory based on variation in the organization of the recognition memory network. Analyses used elastic-net regularized linear regression, random forests, and K-nearest-neighbor regression. However, we did not find that these approaches were effective in predicting performance.

Project 3: Does the network organization of brain regions involved in memory relate to individual differences in memory performance?

Brain regions are usually activated together, and communication between these brain regions is critical to task success. We used several graph theory measures to characterize the organization of brain regions involved in recognition memory. In the same sample of 540 individuals from the Human Connectome Project, we find no relationship between performance and these measures of network organization.

Project 4: How do brain regions involved in social cognition represent individuals and emotions?

The meaning of a social judgment such as “I think that Cory is impressed by Ken”, is determined by the meaning of its parts (“impressed”, “cory”, and “ken”) and the way they are combined (i.e. there is a difference between “Cory is impressed by Ken” and “Ken is impressed by Corey”). We are the first to use neuroimaging to study this conceptual combination mechanism in social cognition.

We examined whether specific brain regions encoded individual features (e.g. ‘ken’, ‘cory’, ‘impressed’) and combinations of features (e.g. “Cory is impressed”, ‘Ken is impressed’). Our methodology is similar to that used by Frankland & Greene, 2020. In addition, we tested which brain regions were involved in representing individuals in their semantic roles (the person doing the thinking, or the target of the emotion).

This work was done in collaboration with Jason Mitchell, and inspired by work from Steven Frankland and Josh Greene on the language of thought hypothesis. Our analyses did not reveal theoretically meaningful effects.

Posted on:
May 23, 2021
3 minute read, 439 words
hugo software
See Also:
Other Memory Studies
Natural Language Processing
Machine Learning