Cross-Participant Decoding of Emotions Elicited Through Nonverbal Vocalizations : Exploring the Use of Searchlight Analysis to Classify Emotions in fMRI Data

Sammanfattning

This thesis analyzed the neural underpinnings involved in the auditory processing of emotional nonverbal vocalizations and studies how these are shared across participants.The main question this thesis aimed to answer is if it is possible to decode the neural signatures of emotions across participants from measures of brain activity and, if so, what brain areas contribute most to this.A sliding local neighborhood decoding approach (searchlight analysis) as well as a whole-brain decoding approach were performed on the fMRI images of 19 participants, where models were trained using the leave-one-participant-out cross validation scheme.The across-participant decoding results showed accuracies close to chance level. Further within-participant analyses showed that within-participant decoding yielded higher accuracies and showed a low inter-participant agreement.These findings suggest that the empathy response to nonverbal emotional vocalizations are hard to generalize across humans. Future work could investigate alternative decoding approaches as well as other experimental designs to further investigate the difficulties associated with the across-participant decoding.