Facial motion engages predictive visual mechanisms

Can deepfakes be used to study emotion perception? A comparison of dynamic face stimuli

Video recordings accurately capture facial expression movements; however, they are difficult for face perception researchers to standardise and manipulate. For this reason, dynamic morphs of photographs are often used, despite their lack of naturalistic facial motion. This study aimed to investigate how humans perceive emotions from faces using real videos and two different approaches to artificially generating dynamic expressions - dynamic morphs, and AI-synthesised deepfakes. Our participants perceived dynamic morphed expressions as less intense when compared with videos (all emotions) and deepfakes (fearful, happy, sad). Videos and deepfakes were perceived similarly. Additionally, they perceived morphed happiness and sadness, but not morphed anger or fear, as less genuine than other formats. Our findings support previous research indicating that social responses to morphed emotions are not representative of those to video recordings. The findings also suggest that deepfakes may offer a more suitable standardized stimulus type compared to morphs. Additionally, qualitative data were collected from participants and analysed using ChatGPT, a large language model. ChatGPT successfully identified themes in the data consistent with those identified by an independent human researcher. According to this analysis, our participants perceived dynamic morphs as less natural compared with videos and deepfakes. That participants perceived deepfakes and videos similarly suggests that deepfakes effectively replicate natural facial movements, making them a promising alternative for face perception research. The study contributes to the growing body of research exploring the usefulness of generative artificial intelligence for advancing the study of human perception.

Stria terminalis, amygdala, and temporoparietal junction networks facilitate efficient emotion processing under expectations

Rapid emotion processing is an ecologically essential ability for survival in social environments in which threatening or advantageous encounters dynamically and rapidly occur. Efficient emotion recognition is subserved by different processes, depending on one's expectations; however, the underlying functional and structural circuitry is still poorly understood. In this study, we delineate brain networks that subserve fast recognition of emotion in situations either congruent or incongruent with prior expectations. For this purpose, we used multimodal neuroimaging and investigated performance on a dynamic emotion perception task. We show that the extended amygdala structural and functional networks relate to speed of emotion processing under threatening conditions. Specifically, increased microstructure of the right stria terminalis, an amygdala white-matter pathway, was related to faster detection of emotion during actual presentation of anger or after cueing anger. Moreover, functional connectivity of right amygdala with limbic regions was related to faster detection of anger congruent with cue, suggesting selective attention to threat. On the contrary, we found that faster detection of anger incongruent with cue engaged the ventral attention "reorienting" network. Faster detection of happiness, in either expectancy context, engaged a widespread frontotemporal-subcortical functional network. These findings shed light on the functional and structural circuitries that facilitate speed of emotion recognition and, for the first time, elucidate a role for the stria terminalis in human emotion processing.

A Novel Dynamic Morphed Stimuli Set to Assess Sensitivity to Identity and Emotion Attributes in Faces

Face-based tasks are used ubiquitously in the study of human perception and cognition. Video-based (dynamic) face stimuli are increasingly utilized by researchers because they have higher ecological validity than static images. However, there are few ready-to-use dynamic stimulus sets currently available to researchers that include non-emotional and non-face control stimuli. This paper outlines the development of three original dynamic stimulus sets: a set of emotional faces (fear and disgust), a set of non-emotional faces, and a set of car animations. Morphing software was employed to vary the intensity of the expression shown and to vary the similarity between actors. Manipulating these dimensions permits us to create tasks of varying difficulty that can be optimized to detect more subtle differences in face-processing ability. Using these new stimuli, two preliminary experiments were conducted to evaluate different aspects of facial identity recognition, emotion recognition, and non-face object discrimination. Results suggest that these five different tasks successfully avoided floor and ceiling effects in a healthy sample. A second experiment found that dynamic versions of the emotional stimuli were recognized more accurately than static versions, both for labeling, and discrimination paradigms. This indicates that, like previous emotion-only stimuli sets, the use of dynamic stimuli confers an advantage over image-based stimuli. These stimuli therefore provide a useful resource for researchers looking to investigate both emotional and non-emotional face-processing using dynamic stimuli. Moreover, these stimuli vary across crucial dimensions (i.e., face similarity and intensity of emotion) which allows researchers to modify task difficulty as required.

Dynamic emotion perception and prior expectancy

Social interactions require the ability to rapidly perceive emotion from various incoming dynamic, multisensory cues. Prior expectations reduce incoming emotional information and direct attention to cues that are aligned with what is expected. Studies to date have investigated the prior expectancy effect using static emotional images, despite the fact that dynamic stimuli would represent greater ecological validity. The objective of the study was to create a novel functional magnetic resonance imaging (fMRI) paradigm to examine the influence of prior expectations on naturalistic emotion perception. For this purpose, we developed a dynamic emotion perception task, which consisted of audio-visual videos that carry emotional information congruent or incongruent with prior expectations. The results show that emotional congruency was associated with activity in prefrontal regions, amygdala, and putamen, whereas emotional incongruency was associated with activity in temporoparietal junction and mid-cingulate gyrus. Supported by the behavioural results, our findings suggest that prior expectations are reinforced after repeated experience and learning, whereas unexpected emotions may rely on fast change detection processes. The results from the current study are compatible with the notion that the ability to automatically detect unexpected changes in complex dynamic environments allows for adaptive behaviours in potentially advantageous or threatening situations.

Fashioning the Face: Sensorimotor Simulation Contributes to Facial Expression Recognition

When we observe a facial expression of emotion, we often mimic it. This automatic mimicry reflects underlying sensorimotor simulation that supports accurate emotion recognition. Why this is so is becoming more obvious: emotions are patterns of expressive, behavioral, physiological, and subjective feeling responses. Activation of one component can therefore automatically activate other components. When people simulate a perceived facial expression, they partially activate the corresponding emotional state in themselves, which provides a basis for inferring the underlying emotion of the expresser. We integrate recent evidence in favor of a role for sensorimotor simulation in emotion recognition. We then connect this account to a domain-general understanding of how sensory information from multiple modalities is integrated to generate perceptual predictions in the brain.