Sound quality impacts dogs' ability to recognize and respond to playback words

Abstract

Humans show a remarkable ability to recognize degraded speech, but they struggle as degradation becomes more severe. Research on dogs, a species naturally exposed to human language, has suggested that also in this species, the level of degradation may negatively correlate with familiar speech recognition. The quality of sound playback equipment is crucial in canine research to ensure accurate reproduction, directly impacting dogs' receptivity. In citizen science, availability, operability, and affordability are also key. However, the varying level of sound degradation of commercially available devices may impact dogs' recognition of the played-back sounds. This study evaluates how different audio devices affect the response to human speech in dogs. Study 1 tested dogs' responses to trained verbal action cues delivered via Augmentative Interspecies Communication (AIC) buttons, loudspeakers, and direct human speech. Study 2 tested Gifted Word Learner (GWL) dogs' accuracy in retrieving named toys using the same devices. Study 3 investigated whether GWL dogs could learn new toy names through recorded voice played via loudspeakers and generalize these to direct human speech. Results consistently showed that dogs' responses to verbal cues were significantly affected by the device type, with the best performance observed for direct human speech and the lowest performance for AIC buttons. Our findings show that the level of sound degradation significantly affects dogs' recognition of recorded speech and suggest that current commercially available AIC devices are not recommended for testing effective recognition of the recorded words.

Keywords: Audio quality; Augmentative interspecies communication (AIC); Dog; Sound processing equipment; Vocal recognition.

Fig. 1

Comparison of the frequency spectrum of the button (yellow, above) and speaker (blue, below) playbacks of white noise (black).

Fig. 2

Action success in dogs when the cue is delivered verbally and by different devices. The error plot shows the proportion of successful responses by dogs to follow the 3 trained action cues in the different experimental conditions: verbal (V), speaker (Sp), speaker with mouth (Sp + M), button (B), and button with mouth (B + M).

Fig. 3

GWL dogs’ success in choosing the requested toy when the toy name is delivered verbally and by different devices. The error plot shows the proportion of successful fetches by GWL dogs when the name of the toy was requested in the different experimental conditions: verbal (V), speaker (Sp), speaker with mouth (Sp + M), button (B), and button with mouth (B + M).

Fig. 4

GWL dogs’ success in choosing the requested toy over 12 experimental trials when the toy name was either trained and then tested verbally (control phase) or using a speaker (treatment phase).