ERP Makers of Valence Coding in Emotional Speech Processing

Abstract

The aim of the study was to investigate brain bioelectrical activity during processing of emotionally valenced speech. At this purpose, 99 meaningful short sentences with a positive emotional valence and other 99 ones with a negative valence, naturally spoken by male and female professional speakers, were selected as such among a larger set of 450 vocal stimuli. 25 neutral sentences containing a first name were also used as target fillers. Stimuli were matched for perceptual properties (e.g., average frequency, intensity, etc..) as well as for phrase’s structure and content (e.g. semantic domain, number of letters and words). All sentences had coherent semantic and prosodic features. 20 healthy participants auditorily listened to stimuli while performing a target detection task consisting in responding to first names. EEG/ERPs were recorded by means of 128 channel recordings.Three event-related potentials (ERP) components (P300, N400 and the Late Positivity) were quantified over anterior scalp areas, and found sensitive to emotional valence as early as 350 ms post-stimulus. The early and late positivities were enhanced by positive content, while N400 was larger to negative content. A similar set of markers was found by previous ERP literature for the processing of positive vs. negative non-verbal vocalizations and music, which hints at a common neural mechanism for extracting the emotional content of auditory information. A swLORETA source reconstruction applied to surface potentials recorded between 350 and 550 ms (P300) showed that while both positive and negative speech activated the superior and middle frontal gyri (indicating semantic working memory), only emotionally negative speech crucially activated the right middle temporal gyrus.