Multivariate fMRI pattern analysis of fear perception across modalities

Eur J Neurosci. 2019 Jun;49(12):1552-1563. doi: 10.1111/ejn.14322. Epub 2019 Jan 23.


The emotional expression of fear can be processed through a number of modalities, and of varying forms, however, much of the functional imaging literature has centered on investigating fear as expressed through faces. Findings point to an active involvement of the amygdala, and remain fairly consistent in other studies of unimodal fear perception; however, few studies have looked at within-subject cross-modal responses to fear. Thus, we approached this inquiry by testing 30 healthy young adults with fast, high-resolution fMRI, recording the neural responses of fear perception, as expressed through faces, bodies, prosody, and vocalizations. The study was analyzed using a multivariate approach (multi-voxel pattern analysis) and yielded a significant distinction in the responses associated with the perception of fearful vs. neutral emotions. Calculated weights highlighted areas in the amygdala and surrounding subcortical structures as contributing the greatest to the discrimination; however, a whole-brain analysis was necessary to obtain above-chance classification accuracy, suggesting that processing fear across modalities likely involves a broad, distributed network. Thus, our findings support a multivariate approach to studying a highly complex construct such as emotion, as it accounts for multiple voxels simultaneously and can accommodate the high subject-level variability that oftentimes comes with studying emotion perception.

Keywords: cross-modality; emotion perception; functional magnetic resonance imaging; multi-voxel pattern analysis.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Brain / diagnostic imaging*
  • Brain / physiology*
  • Brain Mapping / methods*
  • Fear / physiology*
  • Female
  • Humans
  • Magnetic Resonance Imaging / methods*
  • Male
  • Multivariate Analysis
  • Perception / physiology*
  • Young Adult