Mechanisms for Communicating in a Marmoset ‘Cocktail Party’

Abstract

A key challenge for audition is parsing the voice of a single speaker amid a cacophony of other voices known as the Cocktail Party Problem (CPP). Despite its prevalence, relatively little remains known about how our simian cousins resolve the CPP for active, natural communication. Here we employed an innovative, multi-speaker paradigm comprising five computer-generated Virtual Monkeys (VM) whose respective vocal behavior could be systematically varied to construct marmoset cocktail parties and tested the impact of specific acoustic scene manipulations on subjects’ natural conversations. Results indicate that marmosets not only employ auditory mechanisms – including attention – for speaker stream segregation, but also selectively change their own vocal behavior in response to the dynamics of the acoustic scene. These findings suggest notable parallels with human audition to solve the CPP and highlight the active role that individuals play to optimize communicative efficacy in complex real-world acoustic scenes.

A convergent interaction engine: vocal communication among marmoset monkeys

“…As reviewed above, callitrichid turn-taking can be remarkably flexible in various ways. For instance, marmosets discern their partners against a background of other individuals [45,106], or alternate between different affiliative calls depending on the distance from each other [132]. In a follow up study on these findings from the wild, Liao et al [117] could show that directly measured individual arousal levels cannot fully explain when they switch between calls.…”

“…Additional behavioural evidence shows that when multiple individuals are involved in antiphonal conversations, marmosets acutely attend to the calls of each individual separately [44]. In a recent experiment, Jovanovic & Miller [45] exposed marmosets to five playbacks of marmosets emitting phee calls simultaneously. The temporal dynamic of the calls of two pairs of these virtual monkeys suggested engagement in dyadic turn-taking conversations, whereas the fifth was vocally interacting with the focal animal.…”

Optimising source identification from marmoset vocalisations with hierarchical machine learning classifiers

“…However, most studies on marmoset vocal communication are performed under highly controlled conditions, often focusing on a single individual only or physically separated dyads that can interact only vocally with each other. Only rarely are group-level communication dynamics taken into account [27]. Given the highly social nature of marmosets, the next frontier in studying marmoset vocal communication is to do so under more naturalistic conditions in social groups, where they display their communication skills in entirety.…”

The neurobiology of vocal communication in marmosets

“…Given the converging lines of evidence pointing to the role of both the ACx and its various downstream targets in stream segregation and vocalization processing, it is crucial that these experiments be conducted. Partially in service of this goal, recent behavioral studies by Jovanovic and Miller 209 engaged marmoset subjects in an interactive, species‐specific, simulated “cocktail party” scenario in order to better understand marmosets’ behavioral strategies for successfully navigating this core challenge. Coupled with wireless electrophysiology systems, such naturalistic behavioral paradigms offer powerful opportunities to investigate neural mechanisms that underlie the CPP and ASA in the primate brain that have been difficult to study with more conventional approaches.…”

“…209 engaged marmoset subjects in an interactive, species-specific, simulated "cocktail party" scenario…”