Quantifying Physiological Synchrony through Windowed Cross-Correlation Analysis: Statistical and Theoretical Considerations

Abstract

Interpersonal synchrony is a widely studied phenomenon. A great challenge is to statistically capture the dynamics of social interactions with fluctuating levels of synchrony and varying delays between responses of individuals. Windowed Cross-Correlation analysis accounts for both characteristics by segmenting the time series into smaller windows and shifting the segments of two interacting individuals away from each other up to a maximum lag. Despite evidence showing that these parameters affect the estimated synchrony level, there is a lack of guidelines on which parameter configurations to use. The current study aimed to close this knowledge gap by comparing the effect of different parameter configurations on two outcome criteria: i) the ability to distinguish synchrony from pseudosynchrony by means of surrogate data analyses and ii) the sensitivity to detect change in synchrony as measured by the difference between two within-subject conditions. Focusing on physiological synchrony, we performed these analyses on heartrate, skin conductance level, pupil size, and facial expressions data. Results revealed that a range of parameters were able to discriminate synchrony from pseudosynchrony. Window size was more influential than the maximum lag with smaller window sizes showing better discrimination. No clear patterns emerged for the second criterion. Integrating the statistical findings and theoretical considerations regarding the physiological characteristics and biological boundaries of the signals, we provide recommendations for optimizing the parameter settings to the signal of interest.