Perception is subjective; two people may encounter the same physical stimuli (such as an image or sound), but perceive it differently.  This phenomenon is known as perceptual multi-stability--you may be familiar with the popular "blue vs. white dress" and "yanni vs. laurel" debates from recent years on the internet.

​

In my thesis research, we invoked an auditory streaming task in which a listener is presented a sequence of 600 ms tones with the repeated 'ABA-' pattern where 'A' and 'B' are fixed tones but differ from each other in frequency.  For certain frequency differences, a listener may seemingly spontaneously switch between two perceptive states:  1 stream in which the input is perceived as a single galloping-like stimulus or 2 streams in which the input is perceived as two segregated streams, like a Morse code signal.  What neural mechanisms underlie switching between the two perceptual states?

​

We use data-driven methods of time-delayed coordinates, diffusion maps, and dynamic mode decomposition (DMD) to analyze neural recordings (ECoG) from neurosurgical subjects listening to 5-minute long 'ABA-' streams while simultaneously self-reporting switching between perceptual states.

​

Through these data-driven methods, we extract low-dimensional features correlating with both stimulus and perceptual encoding. Importantly, we characterize attractor-like principles from human neural data. Similar mechanisms are proposed in computational models of bistable perception. Our findings support the biological plausibility of such theoretical principles. This work was accepted in the Journal of Neuroscience and is available for [early-release].