Chris Baldassano
LightNews LightNews
banner
chrisbaldassano.bsky.social
Chris Baldassano
@chrisbaldassano.bsky.social
1.3K followers 580 following 26 posts
Associate Professor in Psychology at Columbia, PI of https://www.dpmlab.org/
Posts Replies Media Videos
chrisbaldassano.bsky.social
We also looked at kids' verbal recalls of the movies! Interestingly, we found that recalls in kids who report weaker attachment are more focused on the Searching event in the attachment schema.
Data figure comparing the schematic content of children's recalls based on whether they feel strong or weak attachment to their current caregivers. Children with weak attachment showed significantly greater similarity to the "Searching" schema event
April 10, 2025 at 1:47 PM
chrisbaldassano.bsky.social
Looking at kids’ brain activity related to caregiver stability, we show that kids with unstable caregiver histories have more connectivity between the amygdala and visual processing regions + hippocampus.
Brain map showing regions that are more connected to the amygdala for children with unstable early caregiving, including the hippocampus and lateral temporal regions
April 10, 2025 at 1:47 PM
chrisbaldassano.bsky.social
Our results show that there are indeed differences in brain responses! Kids who report stronger attachment to their current caregiver(s) have more connectivity between the amygdala and vmPFC. Whole-brain results show that heightened amygdala connectivity also shows up in lateral frontal regions.
Data figure showing that amygdala-vmPFC correlation is higher for children with strong attachment, while there is no significant effect on amygdala-dmPFC connectivity
April 10, 2025 at 1:47 PM
chrisbaldassano.bsky.social
We compared brain responses to the movies based on childhood experiences: caregiver stability (caregiver switch/es vs no switch) and caregiver attachment (weak vs strong). We examined response patterns in the amygdala to other regions in the brain with ISFC (Inter-Subject Functional Connectivity).
Figure showing how inter-subject correlation is computed. The average timecourse in the amygdala is computed in one half of the group, and the average timecourse in another brain region such as vmPFC or dmPFC is computed in the other half of the group, and then these two timecourses are correlated.
April 10, 2025 at 1:47 PM
chrisbaldassano.bsky.social
To understand whether childhood experiences such as changing caregivers (in the past) and attachment security (in the present) impact how kids view attachment narratives in movies, we had kids watch a short movie edited to depict 4 crucial events of an attachment schema while collecting fMRI.
Diagram showing the two videos used in the study (Homeward Bound and the Little Princess), and how each video proceeds through four events: "Together" (Characters are together at first, but say goodbye to each other), "Separation" (Characters miss each other because they are still separated), "Searching" (Character searches for other character), and "Reunion" (Characters are reunited and happy to be together again)
April 10, 2025 at 1:47 PM
chrisbaldassano.bsky.social
At NeurIPS this week, led by Caroline Lee in collab w/ Haxby lab: we present the Hyper-HMM, to simultaneously align participants' spatial brain patterns (like hyperalignment) and temporal dynamics (with event segmentation), and align brain events to stimulus features!

www.dpmlab.org/papers/8510_...
December 15, 2023 at 4:30 PM
chrisbaldassano.bsky.social
New preprint led by @huangjiawen.bsky.social: we find that predictable stimuli are better remembered for two separate reasons: making correct predictions improves memory, and likely stimuli can be more easily reconstructed. We can independently manipulate these two factors!

osf.io/preprints/ps...
Image of Figure 2 from the paper, showing that stimuli (moves in a board game called four-in-a-row) are generated based on both move probability from an AI model and participant predictions measured via eye movements.
December 15, 2023 at 3:17 PM
chrisbaldassano.bsky.social
Nice news article on my paper w/ Tal Golan, Matt Siegelman, and Niko Kriegeskorte, describing a method for finding "weak spots" in language models and testing how well they can predict human judgments: bit.ly/3PAbI4w

Full paper is out in Nature Machine Intelligence now! rdcu.be/dl91g

#psychscisky
September 16, 2023 at 9:19 PM