The Sub-Second Dynamics of Spontaneous Mimicry: An Electromyography Study Tracking Infant Caregiver Dyads during Free Play

Spontaneous mimicry (SM) is a ubiquitous feature of human communication (Heyes, 2021; Meltzoff & Williamson, 2017). Research shows that SM is both reflexive and flexible (Wang & Hamilton, 2012). It is sensitive to cues that signify implicit social rules and social hierarchy, suggesting that it is at least partly socially shaped. However, we have yet to map the ontogeny of SM, or its developmental factors (Slaughter, 2021). A marked difference in SM behaviour has been observed in atypical populations (e.g., ASD; Arnold & Winkielman, 2020) increasing the onus for further study. In infants, facial mimicry has been studied extensively and is the central focus of a long running debate surrounding the presence of SM in early infancy (Slaughter, 2021). However, most of these studies have used lab-based tasks or non-naturalistic block-design paradigms (Slaughter, 2021; Meltzoff & Williamson, 2017). The few studies that have observed naturalistic interactions used hand-coded video data, scoring onset and offsets of actions: mimicry was operationalised as an action onset in the observer that occurs within a specific timespan of a prior action onset in the interacting partner (Markodimitraki & Kalpidou, 2019). Here, SM behaviour is gauged in terms of frequency and total number of mimicked actions. They do not measure the magnitude of the action i.e. they cannot record graded changes in action. Employing electromyography (EMG) allows us to decipher moment-to-moment dynamics and sub-second changes.

In the present study, we investigated facial SM behaviour in free-play interactions between 5-months-old infants and their caregivers. EMG electrodes are placed on the facial regions that overly the corrugator supercilii (frowning/eyebrow-movement) in both caregiver and infant. Lab based investigation of SM in infants have found evidence of infant SM of eyebrow movement at this age range (De Klerk et al., 2018). The caregiver and infant were tested while during tabletop free play sessions. The obtained EMG signal is rectified, band-pass filtered and z-scored. Artefacts are rejected by identifying and removing outliers that fall outside of one standard deviation above or below the mean. Cross-correlations are carried out to obtain a comprehensive overview of the temporal correspondence between the partners’ EMG waveforms. Granger causality analyses are also conducted on the EMG-waveforms of the interacting partners to identify if changes in the facial action of one predicts changes in that of the other.

Based on our reading of the literature we had predicted that the cross correlations will be significant when the caregiver’s waveform is lagged (mother mimics the infant) but not when the infant’s waveform is lagged. Our target sample size is 20 dyads, and we are currently at the centriole. In our preliminary analyses (N - 9 dyads), in contrast to our expectations, the cross correlations were significant when the caregiver’s waveform preceded the infant’s (infant’s waveform is lagged). This was seen at lags between .2 and .6 seconds. Granger causality analyses will be performed to test if each of the waveforms can significantly predict the other. Control analysis will be performed with shuffled datasets to rule out spurious results.