Jamal A. Williams

Task-optimized models of sensory uncertainty reproduce human confidence judgments Sensory input is often ambiguous, leading to uncertain interpretations of the external world. Estimates of perceptual uncertainty might be useful in guiding behavior, but it remains unclear whether hu...

New pre-print from our lab, by Lakshmi Govindarajan with help from Sagarika Alavilli, introducing a new type of model for studying sensory uncertainty. www.biorxiv.org/content/10.1...
Here is a summary. (1/n)

I’m also grateful for the support from @PrincetonNeuro.bsky.social, and my D-SPAN award from the NINDS. #neuroskyence #psychscisky #BlackInNeuro #BlackSky

Super grateful for support from my amazing collaborators @ptoncompmemlab.bsky.social, Janice Chen, @chrisbaldassano.bsky.social, Lisa Margulis, and Uri Hasson.

This work also addresses a cinematic puzzle: Why do some movies linger in our minds? Our work shows one way this can happen, by highlighting how repetition of musical themes may help to promote lasting memories for earlier scenes.

By improving our understanding of how music can evoke – and potentially strengthen – naturalistic memories, this work can inform the development of music-based therapies for memory deficits, for example in aging.

In summary, our study bridges the gap between recent work on episodic memory using naturalistic stimuli and decades of work on music-evoked memory retrieval.

Figure showing that the reactivation-memory effect was present across regions in the DMN and the visual peripheral network, but not other cortical networks.

In an exploratory analysis, we checked whether this music-related reactivation extended beyond the DMN. We found specificity to the DMN and visual areas—but not across the whole brain.

Figure showing that the reactivation-memory effect was greater in the music condition than the no-music condition when testing across the full DMN.

As another control, we computed the relationship between reactivation and subsequent memory for the no-music group. When we did this, there was no significant relationship in the DMN; and the music group showed a stronger reactivation–memory link than the no-music group across DMN ROIs.

The findings from this ISC regression analysis suggest that reactivation may have effects on subsequent recall that go beyond the effects of initial encoding strength.

Three figures showing results from our analysis where we regressed out spatial ISC from the original reactivation scores and tested whether the residuals still predicted subsequent memory. Reactivation in PMC still predicted subsequent memory whereas the relationship between reactivation and memory in angular gyrus was no longer significant. The reactivation-memory relationship numerically decreased in angular gyrus but increased in PMC. The effect remained significant in PMC and angular gyrus when performing the analysis as a parcel-wise searchlight.

Yes—the effect was numerically weaker in some areas (angular gyrus) but numerically stronger in other areas (posterior medial cortex); both regions remained significant when using a parcel-based searchlight.

Using ISC from the brain regions that predicted subsequent memory in the music group, we regressed ISC out from their reactivation scores, and we asked: do residual reactivation scores still predict memory?

Illustration of the process of computing spatial inter-subject correlation, showing how one participant’s neural pattern for a scene is correlated with the average of the other participants’ neural patterns for that scene

To tease apart these possibilities, we needed a way of measuring initial encoding strength. Prior studies have found that spatial inter-subject correlation (ISC) predicts subsequent memory; we found this in our study also, validating the use of ISC as a measure of initial encoding strength.

One explanation for this result is that neural reactivation causes improved subsequent recall; alternatively, good initial encoding of a scene could lead to both higher reactivation of the scene and better subsequent recall (leading to a correlation, even if no causal relationship was present).

The top figure is a diagram showing how reactivation scores were sorted based on whether their corresponding scene was remembered or forgotten. The bottom figures show results of the analysis: We found that remembered reactivation scores were higher in key DMN regions; left angular gyrus and left posterior medial cortex. We also found converging evidence when performing the same analysis as a parcel-wise searchlight.

We then split reactivation scores by whether the corresponding scene was later remembered or forgotten.

Main Result: Scenes that were remembered showed significantly higher reactivation in key default mode network (DMN) regions—posterior medial cortex and angular gyrus.

Diagram showing how we compared neural activity from later scenes containing a repeated theme in the music condition to neural activity from earlier scenes containing the same theme in the no-music condition, while controlling for baseline similarity between scenes.

We tracked neural reactivation by comparing activity patterns in the music group (during later music repeats) to patterns in the no-music group (during earlier exposures to the same theme). This gave us a measure of how much non-musical event information was brought back online by the music.

Later on, the same theme plays when Joel and Clementine are out on a lake. Does playing the theme in this scene trigger neural reactivation of the earlier scene, and is the degree of reactivation related to how well participants subsequently remember the earlier scene?

For example, in this scene of the movie, Clementine asks for a Valentine’s Day call and Joel walks away while a musical theme plays…

We also asked whether this reactivation would be associated with better subsequent recall; prior work has shown that retrieval can help memories stick, and we hypothesized that a similar dynamic could be at work here.

The next day, both groups returned and performed spoken free recall for the entire film. We asked: When a musical theme repeated later in the movie, would it reactivate memory for earlier scenes tied to that theme?

We had two groups of participants (n=24 each) watch the full film in the MRI scanner.

🎵 One group saw the movie with the original musical score.
🙉 The other group saw the exact same film with all music removed (dialogue + sound effects intact).

Music is an incredibly powerful retrieval cue. What is the neural basis of music-evoked memory reactivation? And how does this reactivation relate to later memory for the retrieved events? In our new study, we used Eternal Sunshine of the Spotless Mind to find out. www.biorxiv.org/content/10.1...