Neurophysiological correlates of the recognition of facial expressions of emotion as revealed by magnetoencephalography

Marcus R. Streit, Andreas A. Ioannides, Lichan Liu, Wolfgang Wölwer, Jürgen Dammers, Joachim Groß, Wolfgang Gäebel, H.-w. Müller-GärtnerCognitive 1999 · DOI

TL;DR

This study uses magnetoencephalography (MEG) to map brain activity during facial emotion recognition, revealing distinct spatio-temporal patterns compared to simple face recognition. The findings suggest that MEG can specifically identify the neural processes involved in decoding emotional expressions, offering a high-resolution temporal view of this crucial cognitive function.

What Problem Does This Paper Address?

Recognizing facial expressions of emotion is a fundamental human communication skill, developing early in life and thought to be mediated by specialized neural circuitry. While previous studies using PET, fMRI, and EEG have implicated various brain regions (like the inferior frontal cortex, amygdala, and temporal areas), understanding the precise timing and interaction of these areas during emotion recognition has remained a challenge. Existing methods often lack the temporal resolution to disentangle the rapid neural events involved. This paper aims to bridge this gap by using MEG, which offers excellent temporal resolution, to map the spatio-temporal dynamics of brain activity specifically during facial emotion recognition, differentiating it from simpler face perception tasks.

Methods

The core of this study lies in its use of Magnetoencephalography (MEG) to record brain activity. Specifically, they employed a Magnetic Field Tomography (MFT) analysis method. MFT is highlighted as a key innovation here, capable of handling both focal and distributed brain activity without requiring prior assumptions about their form. This allows for more sophisticated analysis of the MEG signals to disentangle spatially distinct, yet temporally overlapping, neural generators.

The experiment involved two main tasks:

  1. Face and Object Recognition (Run 1): Subjects viewed images of faces and various objects, performing a multiple-choice recognition task. This served as a baseline for general face and object processing.
  2. Facial Affect Recognition (Runs 2 & 3): Subjects viewed only faces and were tasked with identifying the specific emotion displayed (happiness, fear, anger, surprise, disgust, sadness).

MEG data was collected using a BTi twin MAGNES system (2 x 37 channels). For one subject, a 148-channel whole-head system was also used in a repeat experiment to assess consistency and capture signals from deeper or more distant brain regions. Data was sampled at 1 kHz, filtered, and averaged. Source localization was performed using MFT, with a focus on identifying regions of high activity (at least 90% of estimated current density) and comparing activations between tasks.

Datasets

The study utilized a dataset of black and white photographs of faces from the Ekman and Friesen set of pictures of facial affect. These were supplemented with photographs of various objects (lorries, chairs, horse heads, birds, and flowers) for the control task. The stimuli were digitized, adjusted for luminance, and presented in a randomized order. No new datasets were introduced; existing, well-established stimuli were used.

Key Insights and Ideas

  • MEG's Temporal Precision is Key: The study strongly supports the idea that MEG, with its millisecond-level temporal resolution, is crucial for understanding the dynamic, sequential nature of brain processing for complex cognitive tasks like emotion recognition.
  • Distinct Neural Signatures for Emotion vs. Simple Face Recognition: The most significant finding is that the brain's response to faces differs markedly depending on whether the task demands simple recognition or the identification of emotional expression. This suggests specialized neural pathways are engaged.
  • A Sequential Activation Pattern: The research identified a roughly chronological sequence of brain activations during facial emotion recognition, starting in the posterior superior temporal cortex, moving to the middle temporal cortex and amygdala, and finally involving the left inferior frontal cortex and anterior cingulate gyrus.
  • The Amygdala's Role in Emotion: The amygdala showed selective activity during facial expression judgment, particularly in some subjects, reinforcing its known role in processing emotional significance.
  • Right Hemisphere Advantage (Potentially): The findings suggest a potential right hemisphere advantage for processing emotional expressions, with earlier and stronger activations observed in right temporal regions.
  • Whole-Head System Enhances Deep Source Detection: The comparison between the twin-channel and whole-head systems highlighted the latter's superior ability to detect activity from deeper brain structures like the amygdala and anterior cingulate gyrus.

Key Results

  • Differential Activation: Facial emotion recognition activated specific brain areas (e.g., posterior sector of right superior temporal cortex, right inferior occipitotemporal regions, middle temporal cortex bilaterally, amygdala, left inferior frontal cortex, right anterior cingulate gyrus) that were either not activated or showed significantly lower activation during simple face recognition.
  • Temporal Dynamics:
    • The posterior sector of the right superior temporal cortex was among the first areas to show significantly stronger activation for affect judgment (140-170 ms).
  • The middle sector of the right temporal cortex showed activation around 180-210 ms, followed by a reactivation in all subjects around 210-240 ms.
  • The left inferior frontal cortex showed activity later, around 230-300 ms.
  • The amygdala showed activity around 220 ms in some subjects.
  • Behavioral Performance: Subjects performed well above chance on the emotion recognition task (60-93% accuracy), indicating they understood and executed the task correctly.

Limitations and Future Work

  • Small Sample Size: The study acknowledges the limitation of a small sample size (four healthy male volunteers), particularly for the whole-head system experiment. This means conclusions are preliminary and require further validation with larger, more diverse cohorts.
  • Reliance on Averaged Signals: The study primarily relied on averaged MEG signals, which can mask trial-to-trial variability and single-trial dynamics. The authors mention they are currently processing the data using single-trial analysis to investigate individual emotional expressions and their specific neural correlates.
  • Control for Cognitive Demands: Future studies should include additional tasks to control for cognitive demands like working memory, which might influence the observed activations.
  • Further Investigation of Interplay: The authors emphasize the need for further studies to investigate the interplay between brain regions and to determine whether their interpretation of spatio-temporal activation patterns holds true across different tasks and individuals.

Summary updated

BibTeX
@inproceedings{streit1999neurophysiological,
  author = {Marcus R. Streit and Andreas A. Ioannides and Lichan Liu and Wolfgang Wölwer and Jürgen Dammers and Joachim Groß and Wolfgang Gäebel and H.-W. Müller-Gärtner},
  title = {{Neurophysiological correlates of the recognition of facial expressions of emotion as revealed by magnetoencephalography}},
  booktitle = {Cognitive},
  year = {1999},
  doi = {10.1016/s0926-6410(98)00048-2},
}
Figure