Audiovisual Representations of Valence: a Cross-study Perspective

Area of Science:

• Neuroscience
• Psychology
• Cognitive Science

Background:

• Hedonic valence, the pleasantness or unpleasantness of psychological states, is a core component of emotional experience.
• Multivariate pattern analysis (MVPA) of functional magnetic resonance imaging (fMRI) data allows for the identification of valence from distributed brain activity patterns.
• A key challenge in MVPA research is ensuring construct validity, as classification results may be influenced by study-specific factors rather than valence itself.

Purpose of the Study:

• To determine if hedonic valence can be identified across different fMRI studies with varying experimental parameters.
• To enhance the construct validity and generalizability of findings related to valence representation in the brain.
• To investigate the robustness of valence classification across diverse stimuli and experimental paradigms.

Main Methods:

• A leave-one-study-out cross-validation approach was employed, training classifiers on data from five fMRI studies and testing on the sixth.
• Six independent fMRI studies utilizing auditory, visual, or audiovisual stimuli were combined, totaling 93 participants.
• Whole-brain and searchlight analyses were performed on fMRI data to classify positive versus negative valence states.

Main Results:

• Whole-brain classification achieved 72% accuracy in distinguishing between positive and negative valence states across studies.
• Searchlight analysis localized valence representation to specific brain regions, including the right postcentral and supramarginal gyri, left frontal cortices, and medial frontal cortices.
• The results demonstrate reliable cross-study classification of hedonic valence, indicating generalizable neural representations.

Conclusions:

• Hedonic valence can be reliably classified across diverse fMRI studies, supporting its generalizable neural representation.
• The findings enhance the construct validity of using fMRI-based MVPA to study emotional valence.
• This approach provides a robust method for investigating the neural underpinnings of emotional experience across different contexts.