Our research examines how memory emerges from interactions between cognition, behavior, and brain activity. We combine computational models with behavioral experiments and EEG/fMRI analyses to test mechanistic theories rather than only describe neural correlates. This approach helps us explain classic memory effects, develop new analysis methods, and constrain formal models with converging evidence. To explore these ideas directly, try the interactive demos below, including walkthroughs of EEG signals, BOSC oscillation detection, and memory modeling effects.

Electroencephalography (EEG) records the brain's electrical activity through electrodes on the scalp. Two major analysis approaches — Event-Related Potentials (ERPs) and oscillations — reveal different aspects of brain function. Explore the interactive demo below to build an intuitive understanding of how EEG works.

Brain signals contain rhythmic activity (oscillations) that are thought to play key roles in cognition. But detecting these oscillations is surprisingly tricky: brain signals naturally have more power at low frequencies (a "1/f" pattern), so naive methods are biased toward finding oscillations at low frequencies even when none exist. BOSC solves this. Explore the interactive demo below to see how.

Item memory refers to the ability to recognize individual items — judging whether a word, face, or object has been encountered before, independent of when or where it was studied. Electroencephalography (EEG) can be used to study item memory by examining Event-Related Potentials (ERPs) — voltage changes in the brain time-locked to a stimulus. In recognition memory research, three effects have proven especially informative: the Subsequent Memory Effect (brain activity at study predicts later remembering), the Old/New Effect (brain activity at test distinguishes correctly recognized old items from correctly rejected new ones), and the Retrieval Success Effect (brain activity at test distinguishes successfully from unsuccessfully retrieved old items). Explore the interactive demo below to see how these effects look and what cognitive processes they reflect.

Can we actually predict whether a memory will form, rather than just describe brain activity after the fact? Chakravarty, Chen & Caplan (2020) applied machine learning classifiers to study-phase EEG to answer this question. By mapping EEG features — such as the Late Positive Component (LPC) and Slow Wave amplitudes — to later memory outcomes, classifiers can predict hit or miss trial-by-trial. This has practical implications for adaptive learning systems, clinical diagnostics, and basic neuroscience. Explore the interactive demo below to understand how classifiers work, why single-trial prediction is challenging, and what the ROC curve and AUC tell us about classifier performance.

Why does producing words aloud (writing, saying) improve memory compared to silent reading, and why does this production effect create list-strength effects — while simply studying items longer does not? Attentional Subsetting Theory proposes that attention encodes only a small subset of a word's features, and the size of the relevant feature space determines whether items interfere with each other in memory. Explore the interactive demo below to see how.

In recognition memory, stronger study does not always help uniformly: under some mixed-list conditions, weaker items can be remembered better than expected, producing the Inverted List-Strength Effect. This interactive demo walks you through a short study-test cycle so you can observe how item strength, interference, and retrieval dynamics combine to create this counterintuitive pattern.