Comparing bivariate and multivariate timeseries analysis in joint action using cross-recurrence quantification analysis

When pursuing a shared goal, pairs of individuals act in ways that reflect the reciprocal relationships between individual and interpersonal capabilities, demands, and actions. One important question facing researchers is how to best analyze these joint action data given the many behaviors spread out across multiple actors that contribute to achieving the shared outcome. In this paper, we compare the analysis of interpersonal motor coordination when using a single measured timeseries from each actor to using multivariate (more than two) timeseries when using cross-recurrence quantification analysis (CRQA). Pairs of participants completed a joint Fitts’s task by moving their arms between two targets relative to one another (in-phase or anti-phase). Asymmetries in the task demands were produced by varying the relative distances participants had to move between targets and individual stance demands. Our results indicate that when using a multivariate timeseries from each actor for phase space reconstruction, CRQA was more sensitive to changes in coordination dynamics brought about by these experimental manipulations, suggesting that when available, joint action researchers would benefit from using multivariate timeseries in the analysis of behavior.