Fig 1.
In A) a joint finger tapping paradigm is illustrated. Two persons (dyad members) tap an isochronous rhythm together. Their auditory feedback can be manipulated so that the dyad is bidirectionally coupled, i.e. that dyad member 1 hears dyad member 2 and vice versa, here illustrated in the top right. In the two bottom illustrations we see unidirectional coupling, wherein the information between dyad members only goes one way, for instance so that dyad member 1 only hears dyad member 2, and dyad member 2 hears only themselves. B) Time series representing the intertap interval, a measure of the time between successive taps, of each dyad member. Colours indicate dyad member. When these time series are cross-correlated at lag -1, lag 0, and lag +1, a pattern such as illustrated in C) emerges. Here, the pattern would indicate a mutual adaptation synchronization strategy.
Fig 2.
In A we see the four-oscillator model, with the oscillators represented as circles within the two units. The coupling terms are shown as arrowed lines. In B the coupling matrix Knp is shown, and two out of 12 significantly different lag patterns produced by the model are shown in C.
Fig 3.
In the first row, synchronization patterns from the empirical data are shown. Leading-leading (1) corresponds to a subgroup from dataset 1. Leading-following (2) and (3) are from the two unidirectional conditions in dataset 2. From the mutual adaptation group, (4) is the bidirectional condition from dataset 2, whereas (5) and 6) are the two remaining subgroups from dataset 1. The green lag patterns show the empirical data. The blue lag patterns show the synchronization strategy patterns produced by the model at the given coupling weights listed in Table 1. The patterns are plotted as the mean value, with error bars indicating the standard error of the mean. Note that the lag 0 component in the simulated data is consistently more positive than in the empirical data, likely as a result of the continuous coupling in the model.
Table 1.
Overview of the best coupling weights found for each group.
The numbering of the groups corresponds to the labelling used in Fig 3. The Bhattacharyya coefficient listed here is the mean coefficient between the three lags. The coupling strengths i1, e1, i2, e2, from the coupling matrix K are measured in 1/s and scaled according to the integration time step used in the simulations. For details, see Methods section.
Fig 4.
Illustration of regions involved in interpersonal synchronization.
Motor regions (shown in light blue) are bidirectionally linked to auditory regions (shown in light yellow), and with the temporoparietal junction (shown in light red). Actions produced by one individual’s motor system is perceived in auditory regions of the other individual.
Fig 5.
Synchronization as measured by the synchronization index as a function of coupling weight.
In A we see the synchronization index of the two-oscillator model as a function of coupling weight. The vertical orange line indicates the point of maximum synchronization. In B the same is shown for the four-oscillator model.
Fig 6.
Clustering dendrogram and resulting lag patterns.
In A and in B the clustering dendrogram for, respectively, the two-oscillator model and the four-oscillator model are shown. In C the corresponding mean lag patterns for the two-oscillator model is shown. In this case, the two-oscillator model produced six significantly different lag patterns. Out of these six, three exhibit a strong lag-0 component (clusters 1, 5, and 6), and the remaining three shows weak correlations across the lags. In D we show the same procedure applied to data from the four-oscillator model. Here we see a much richer variety of lag patterns, with 12 being significantly different. Note that the lag patterns produced from the clustering algorithm do not necessarily constitute synchronization strategies, but rather patterns that are quantitatively different from each other. The high number of different clusters also, in part, stems from the flip symmetry in the system, as is seen for instance in cluster 1 and cluster 8.