Multiscale kinematic analysis reveals structural properties of change in evolving manual languages in the lab

Reverse engineering how language emerged is a daunting interdisciplinary project. Experimental cognitive science has contributed to this effort by eliciting in the lab constraints likely playing a role for language emergence; constraints such as iterated transmission of communicative tokens between agents. Since such constraints played out over long phylogenetic time and involved vast populations, a crucial challenge for iterated language learning paradigms is to extend its limits. In the current approach we perform a multiscale quantification of kinematic changes of an evolving silent gesture system. Silent gestures consist of complex multi-articulatory movement that have so far proven elusive to quantify in a structural and reproducable way, and is primarily studied through human coders meticulously interpreting the referential content of gestures. Here we reanalyzed video data from a silent gesture iterated learning experiment (Motamedi et al. 2019), which originally showed increases in systematicity of gestural form over language transmissions. We applied a signal-based approach, first utilizing computer vision techniques to quantify kinematics from videodata. Then we performed a multiscale kinematic analysis showing that over generations of language users, silent gestures became more efficient and less complex in their kinematics. We further detect systematicity of the communicative tokens’s interrelations which proved itself as a proxy of systematicity obtained via human observation data. Thus we demonstrate the potential for a signal-based approach of language evolution in complex multi-articulatory gestures