Taft Lectures

Brains are dynamic networks. Think of the brain as a city, where regions are neighborhoods and connectivity forms the roads. Real brain function is traffic: communication routes reconfigure over time and differ across individuals. Understanding this heterogeneity requires statistical methods that can handle high-dimensional, correlated data, distinguish signal from noise, and quantify uncertainty about change.

Statistical methods are indispensable in this setting. They provide a principled way to handle high-dimensional dependence, avoid mistaking noise for structure, and quantify uncertainty in claims about when and how the brain changes. These steps are crucial if connectivity is to become a reliable marker of individual brain functioning.

In this talk, I will describe modern statistical methods for quantifying how brain activity and connectivity vary over time and across individuals, and why that variation matters. I will focus on (i) multi-subject analyses that discover subgroups with similar activity and connectivity patterns, especially when integrated with subject-level information; (ii) unified approaches to dynamic functional connectivity that characterize time-varying network interactions directly from multivariate fMRI time series; and (iii) models that link imaging and non-imaging predictors to behavioral or clinical outcomes. To emphasize that these ideas extend across scales, I will also point to recent results in calcium imaging, where decoding noisy fluorescence recordings identifies coherent neuronal ensembles and how their coordination changes with spatial context during navigation.

The long-term goal is practical: better explanations and predictions, and more targeted strategies for clinical screening and intervention.

Refreshments will be served 3:15-3:45pm in the Math Faculty & Graduate Student Lounge Room 4118 French Hall West