Dimensions of Neural Reliability


Jacek Dmochowski

(Stanford University)

Please LOG IN to view the video.

Date: December 4, 2013


Reliability of neural signals was first investigated by Hasson et al. (1), who introduced the inter-subject correlation (ISC) measure as a model-free technique to identify brain areas which respond stereotypically to natural stimuli (i.e., film clips). In this talk I will present a series of experiments which illustrate the utility of neural signal reliability in both naturalistic and controlled behavioral studies. At the core of the findings is a new signal decomposition technique which extracts components of the data exhibiting maximal correlation across repetitions of the stimulus. I will show that time-resolved correlations in the space of these components may serve as a readout of viewer engagement (2). Next, I will try to argue that neural reliability observed across a small group of viewers may be used to predict the behavioral response of a much larger audience. Finally, I will demonstrate the application of reliability analysis to conventional ERP data sets, illustrating with a fine motion discrimination paradigm that “correlated components analysis” yields a high-SNR dimensionality reduction, facilitating studies of perceptual processes.

Further Information:

(1) Hasson, Uri, et al. “Intersubject synchronization of cortical activity during natural vision.” Science 303.5664 (2004): 1634-1640. (2) Dmochowski, Jacek P., et al. “Correlated components of ongoing EEG point to emotionally laden attention–a possible marker of engagement?.” Frontiers in human neuroscience 6 (2012).

Created: Wednesday, December 4th, 2013