Dynamic vocal learning in adult marmoset monkeys

Socially mediated vocal plasticity, such as vocal accommodation, is an important mechanism for maintaining social cohesion and coordination in group-living animals. Yet, the cognitive mechanisms underlying vocal accommodation remain poorly understood. Here we scrutinize the temporal dynamics of vocal accommodation in common marmosets by tracking the vocalizations of pairs during the first two months after pair formation. We show that vocal accommodation in marmoset pairs displayed the following properties: (i) bidirectional learning (i.e. both males and females underwent equal amount of vocal change); (ii) an exponential decrease in vocal distance with time; (iii) sensitivity to initial vocal distance, that is, pairs converged to a higher extent when their vocalizations were initially more dissimilar; and (iv) synchronous dyadic movement through acoustic space. We find that mathematical models capturing these characteristics and accurately simulating the phenomena rely on dynamic template matching. This suggests that marmosets match their vocalizations to an auditory template continuously updated to track changes in partners' vocalizations, a dynamic form of vocal learning. The model thus suggests that vocal convergence is a highly interactive process that may have provided a foundation for more advanced vocal learning in primates, resonating with recent claims about its role in language evolution.