Oscillatory brain dynamics underlying affective face processing

Abstract

Facial expressions are ubiquitous and highly reliable social cues. Decades of research has shown that affective faces undergo facilitated processing across a distributed brain network. However, few studies have examined the multispectral brain dynamics underlying affective face processing, which is surprising given the multiple brain regions and rapid temporal dynamics thought to be involved. Herein, we used magnetoencephalography to derive dynamic functional maps of angry, neutral, and happy face processing in healthy adults. We found stronger theta oscillations shortly after the onset of affective relative to neutral faces (0-250 ms), within distributed ventral visual and parietal cortices, and the anterior hippocampus. Early gamma oscillations (100-275 ms) were strongest for angry faces in the inferior parietal lobule, temporoparietal junction, and presupplementary motor cortex. Finally, beta oscillations (175-575 ms) were stronger for neutral relative to affective expressions in the middle occipital and fusiform cortex. These results are consistent with the literature in regard to the critical brain regions, and delineate a distributed network where multispectral oscillatory dynamics support affective face processing through the rapid merging of low-level visual inputs to interpret the emotional meaning of each facial expression.

Keywords: MEG; affective faces; beta desynchronization; gamma activity; magnetoencephalography; theta activity.

Figure 1.

Task design. Face stimuli were presented centrally for 1500 ms and preceded by a fixation cross for a random duration between 1500 and 1700 ms. On each trial, participants were instructed to identify each face as either male or female using a button pad.

Figure 2.

Behavioural results: reaction time differences across expressions. Participants exhibited slower responses during angry compared to neutral face trials. The dots represent the mean reaction time for each participant per affective expression. The box plots illustrate the mean, first, and third quartiles, and the whiskers indicate the minima and maxima. The violin plots illustrate the probability density. Horizontal bars depict the significant pairwise differences (* P < .01, Bonferroni corrected).

Figure 3.

Sensor- and source-level oscillatory activity. (Left) Time-frequency spectrograms illustrate the oscillatory responses across all trials and participants from three representative sensors. Time (ms) is shown on the x-axis and frequency (Hz) on the y-axis, and the colour scale illustrates the change in oscillatory power relative to the baseline period. Strong increases in gamma (top; MEG1922, posterior sensor near occipital cortices) and theta (bottom; MEG2113, posterior sensor near occipital cortices) were observed, in addition to alpha and beta decreases (middle; MEG1623, sensor near left parietal cortex) from baseline. (Right) 3D renditions illustrate the mean images (pseudo-t; see colour bar) for each oscillatory response (i.e. time-frequency window).

Figure 4.

Differences in theta oscillatory activity between angry, neutral, and happy faces. In each panel, the brain images illustrate the F-values for facial affect differences thresholded at P < .005, corrected. In the data plots, the dots represent the relative oscillatory power for each participant per condition, taken from the peak-voxel in the corresponding clusters. The box plots show the mean, first, and third quartiles, and the whiskers indicate the minima and maxima. The violin plots illustrate the probability density. Horizontal bars depict the significant pairwise differences (** P < .001, * P < .01, all Bonferroni corrected).

Figure 5.

Differences in oscillatory activity between angry, neutral, and happy faces for (A) beta and (B) gamma responses. The brain images illustrate the F-values for facial affect differences thresholded at P < .005, corrected. In the data plots, the dots represent the relative oscillatory power for each participant, taken from the peak-voxel in the corresponding clusters. The box plots illustrate the mean, first, and third quartiles, and the whiskers indicate the minima and maxima. The violin plots illustrate the probability density. Horizontal bars depict the significant pairwise differences (** P < .001, * P < .01, all Bonferroni corrected).