Nerual underpinnings of perceptual biases
Humans exhibit an attractive bias towards previously seen stimuli known as serial dependence. In this project, I used functional MRI to look at visual brain activity in a task where individuals showed this bias. I applied decoding techniques to get single trial estimates of the stimulus in different brain regions, and found (surprisingly) that these representations were repelled from previous stimuli. To explain these divergent behavioral and neural biases, I used a computational model to demonstrate that readout from early sensory areas is continuously monitoring and accounting for sensory adaptation (and that such an arraignment is energy efficient).
Tools used: behavioral psychophysics, fMRI, linear decoding, Bayesian/computational modeling
Neural Signal complexity predicts memroy performance
In this project, I was initially tasked with predicting whether an individual would subsequently remember a stimulus based on spectral features of their intracranial EEG signals. I noticed that the patterns picked up by the classifier were distinct for high and low performing subjects. A subsequent exploratory analysis revealed a strong relationship between neural signal complexity (indexed by spectral 'slope' or sample entropy) and task performance. This pattern even extended to the level of single trials (eg. more complexity on subsequently remembered trials) suggesting more complex/ information rich states are optimal for encoding new associations.