Multivariate pattern classification reveals autonomic and experiential representations of discrete emotions.

Published

Journal Article

Defining the structural organization of emotions is a central unresolved question in affective science. In particular, the extent to which autonomic nervous system activity signifies distinct affective states remains controversial. Most prior research on this topic has used univariate statistical approaches in attempts to classify emotions from psychophysiological data. In the present study, electrodermal, cardiac, respiratory, and gastric activity, as well as self-report measures were taken from healthy subjects during the experience of fear, anger, sadness, surprise, contentment, and amusement in response to film and music clips. Information pertaining to affective states present in these response patterns was analyzed using multivariate pattern classification techniques. Overall accuracy for classifying distinct affective states was 58.0% for autonomic measures and 88.2% for self-report measures, both of which were significantly above chance. Further, examining the error distribution of classifiers revealed that the dimensions of valence and arousal selectively contributed to decoding emotional states from self-report, whereas a categorical configuration of affective space was evident in both self-report and autonomic measures. Taken together, these findings extend recent multivariate approaches to study emotion and indicate that pattern classification tools may improve upon univariate approaches to reveal the underlying structure of emotional experience and physiological expression.

Full Text

Duke Authors

Cited Authors

  • Kragel, PA; Labar, KS

Published Date

  • August 2013

Published In

Volume / Issue

  • 13 / 4

Start / End Page

  • 681 - 690

PubMed ID

  • 23527508

Pubmed Central ID

  • 23527508

Electronic International Standard Serial Number (EISSN)

  • 1931-1516

International Standard Serial Number (ISSN)

  • 1528-3542

Digital Object Identifier (DOI)

  • 10.1037/a0031820

Language

  • eng