Stress-induced flexibility and individuality in female and male zebra finch distance calls

Abstract

Vocal recognition is central to the coordination and organization of behavior in pair-bonding species such as zebra finches. Zebra finches' vocalizations are individualized and support acoustic discrimination processes. Physiological states - such as the ones involved in emotional stress - can modify vocal production and consequently the structure of vocalizations. These modifications might signal the state of the caller but also impair individual recognition processes. This may represent a signaling trade-off, especially in contexts where both pieces of information can be critically important, for example when mates use calls to reunite after social isolation. Here we study the impact of a stress on the individual vocal signature in both female and male zebra finch distance calls. We built a manually curated database of distance calls of several individuals (both females and males) recorded in control and stress conditions. The stress was induced either by social isolation of the bird or using exogenous corticosterone. We developed a machine learning approach to assess the impact of stress on the individual characterization of calls. We show that while calls' spectral structure is significantly modified by stress, it still allows for the correct classification of calls to the caller. Moreover, we also show that the stress-induced modification of calls' structure is not a 'general feature signal' that can be detected as a 'stress' signal regardless of identity. Thus, female and male zebra finch calls' structure show stress-induced flexibility that stays within the range of individual vocal signatures.

Keywords: Individual signature; Machine learning; Stress; Zebra finch.