Functional Connectivity of Auditory Statistical Learning

Abstract

Statistical learning is a key mechanism underpinning language acquisition in infancy. Using electrophysiological functional imaging, researchers have demonstrated neural adaptation to statistical probabilities in a speech stream using event-related paradigms, but the role of functional connectivity is unknown. The current study investigates the functional connectivity of statistical learning using magnetoencephalography (MEG) recordings of adults and adolescents listening to a speech stream composed of tri-syllabic pseudowords. Data were source localized using the eLORETA inverse method and band-pass filtered into canonical bands, theta (4-7 Hz), alpha (8-12 Hz), low beta (13-18 Hz), high beta (19-30 Hz), and low gamma (31-55 Hz). Two methods of connectivity estimation are compared, amplitude envelope correlations, which are sensitive to slow modulations in cortical activity, and the instantaneous amplitude correlation, which has a higher temporal resolution. Functional networks are derived and characterized using both methods. Support vector regression is used to assess whether network features predict statistical learning. Results of the investigation show that high temporal resolution metrics predict auditory statistical learning above chance, independently of event-related measures of neural adaptation.