Individual level recognition of familiar human speakers in dogs

Abstract

Voice-based individual level recognition facilitates efficient navigation in the social environment for vocal species. Previous studies have indicated that certain species responded differently to calls of different individuals. Investigating individual level recognition in animals is, however, challenging, as distinguishing it from class level recognition (e.g. familiar/unfamiliar) is difficult. Previous studies showed that dogs, Canis familiaris, can choose their owner's voice from among strangers' voices, but whether this performance reflects class or individual level recognition remained unclear. Here, dogs had to choose the vocalizer from among three familiar speakers (owners) based on prerecorded speech in a cross-modal match-to-sample task. We predicted that if dogs are capable of voice-based individual level recognition, they will successfully match the voice to the vocalizer. Dogs' behaviour and its association with acoustic distance between speakers were analysed. Our results showed that, on the group level, dogs' choosing success and the proportion of the time dogs spent looking at the target owner were significantly above chance. Dogs' performance, however, was not significantly correlated with speakers' acoustic distance in either of the measured cues. The performance of choosing the speaker from among familiar people provides evidence for individual level recognition in dogs. Further research will have to reveal the contributions of specific acoustic cues to this performance.