Decoding emotions from brain activity and connectivity patterns

Research output: ThesisDoctoral ThesisCollection of Articles


Emotions guide both human and animal behavior providing the means for survival in a constantly changing environment. Different emotions seem to be distinct from each other in several aspects, including physiological changes, bodily sensations, facial expressions, and subjective experience. Whether and how such emotion categories exist at the neural level remains however under debate. The goal of this dissertation was to employ pattern classification methods to investigate the neural underpinnings of different emotion states. Specifically, it was hypothesized that if different emotions have distinct neural bases, we should be able to reliably classify them from brain activity and connectivity patterns. Further, it was hypothesized that the classifier confusions presumably reveal which emotions have similar neural substrates. Multiple emotional states were induced in four studies with altogether 109 participants using emotional movies, mental imagery, and narratives while participants' brain activity was measured with functional magnetic resonance imaging (fMRI). Several approaches to the fMRI data analyses were employed: multivariate pattern classification to distinguish voxel activity and functional connectivity patterns underlying different emotions, representational similarity analysis to compare experienced and neural similarity of different emotions, functional connectivity analysis to reveal emotional modulations in brain connectivity, univariate methods such as general linear model (GLM) to visualize the neural substrates of different emotions, and correlation analyses to compare the relationship of different emotions at different emotion-related components. Results from these studies show that specific emotions can be classified from both voxel activity and functional connectivity patterns. Successful pattern classification of voxel activity across the whole brain shows that different emotions have distinct brain activity patterns that generalize across participants and across emotion induction techniques. Further, emotions that subjectively feel more similar also have more similar neural underpinnings. Functional connectivity is modulated by emotional content and shows distinct patterns for different emotions especially within the default mode network (DMN). DMN regions especially in the cortical midline, together with somatomotor, sensory, and subcortical areas, support most emotions. Finally, distinctness of emotions is related at the level of different components, including facial expressions, bodily sensations, emotional evaluations, subjective experiences, and neural substrates. To conclude, emotions have distinct brain activity and connectivity patterns that encompass large extent of the brain. Emotions can thus be viewed as systemic states that, at a given moment, facilitate and constrain other mental functions.
Translated title of the contributionTunteiden luokittelu aivojen aktivaatiosta ja konnektiviteetista
Original languageEnglish
QualificationDoctor's degree
Awarding Institution
  • Aalto University
  • Sams, Mikko, Supervising Professor
  • Nummenmaa, Lauri, Thesis Advisor
  • Jääskeläinen, Iiro, Thesis Advisor
Print ISBNs978-952-60-7817-5
Electronic ISBNs978-952-60-7818-2
Publication statusPublished - 2018
MoE publication typeG5 Doctoral dissertation (article)


  • emotion
  • brain
  • pattern classification
  • functional connectivity
  • fMRI
  • MVPA
  • RSA


Dive into the research topics of 'Decoding emotions from brain activity and connectivity patterns'. Together they form a unique fingerprint.

Cite this