Spatiotemporal dynamics and substates underlie emotional signalling in facial movements.

From overt emotional displays to a subtle eyebrow raise during speech, facial expressions are key cues for social interaction. How these inherently dynamic facial signals encode emotion across non-verbal expression and speech remains only partially understood. In Study 1 we recorded participants’ facial movements signalling happy, sad and angry emotions in Expression-only and Emotive-speech conditions. We employed a data-driven pipeline integrating facial motion quantification, spatiotemporal classification and clustering to investigate the structure and function of facial dynamics in signalling emotion. Results reveal that a few spatiotemporal patterns reliably differentiated emotion in non-verbal expressions and emotive speech facial signals. Furthermore, we identified transient substates – or dynamic phases – that are diagnostic of emotion intent and conditions. A perceptual validation with naïve observers (Study 2) showed that the low-dimensional spatiotemporal structure captures meaningful cues that closely predict human emotion categorisations. We discuss theoretical implications of a low-dimensional spatiotemporal structure for optimal transmission and perception of dynamic facial emotion signals and face-to-face interaction. This work also provides a framework for modelling dynamic social cues and insights for the design of expressive emotive capabilities in social agents.