Selective attention enables separation of overlapping speech in noisy environments. This study using EEG-fMRI fusion with continuous audiovisual cocktail party speech, by Patrik Wikman, Viljami Salmela and co-authors, shows that attention acts by routing neural processing through recurrent feedback-feedforward loops between nodes of the speech network.
Selective attention to speech is often studied using highly simplified and artificial settings. However, this fMRI study by Artturi Ylinen and colleagues demonstrates that properties of the task performed with regard to the attended speech (e.g., focus on meaning or focus on phonological content) affect neural activations in widespread regions of the brain. This indicates that naturalistic tasks are likely required, if one wishes to understand neural processing as it occurs in everyday life.