A new study published in Scientific Reports questions the functionality of using the Augmentative Interspecies Communication (AIC) button with dogs. This tool has captivated social media and fueled a global citizen science movement. Researchers from the Department of Ethology at Eötvös Loránd University in Budapest, Hungary, reveal that audio quality significantly influences a dog's ability to recognize and respond to recorded words.
While dogs recognize recorded speech to some extent, they struggle more as the degradation becomes more severe, as is the case with the AIC buttons. Researchers examined how dogs respond to familiar words using two types of audio recordings and playback devices: AIC buttons and a smartphone-loudspeaker configuration, each exhibiting different levels of sound degradation.
The study tested 17 dogs with trained action verbal cues such as "lie down," "spin," and "raise a paw" through the two playback devices and direct human speech. Dogs excelled at responding to direct human speech, achieving nearly 100% success, and their success was around 70% when using loudspeaker playbacks. However, dogs' correct responses declined sharply to only 30% of correct trials with AIC buttons.
Similar results were observed for GWL dogs, known for their extensive vocabulary, when they were tested with the same devices to identify familiar toys. "The frequency analysis showed that the loudspeaker preserves most sound frequencies of the human speech, the AIC button loses frequencies that would be necessary to convey human speech to dogs," explains the coauthor, Dr. Tamás Farago.
The incredible ability of GWL dogs to learn toy names rapidly prompted the researchers to give these dogs an additional challenge. GWL-dog owners were asked to teach new toy names with recordings instead of using their voice. Only the loudspeakers were utilized for this study. When teaching, owners engaged in their usual social, playful interactions with the toy and the dogs while saying – e.g., "Look! This is…", but then, instead of saying the novel toy name, they tapped the play icon on their smartphone which emitted the pre-recorded toy name through the speaker. After two weeks of teaching, at the test, the owner played a recording of a toy's name and asked the dogs to fetch it from a selection of toys in another room.
GWL dogs were able to learn new toy names through playback from speakers; however, their performance was around 70%. Next, the scientists tested whether the dogs could recognize the toy name learned only from the recordings when they heard this as spoken directly by the owners. GWL dogs performed slightly better in these direct verbal tests, despite having learnt from recorded names during their training. This suggests that GWL dogs can transfer what they learned from recordings – which imply some level of sound degradation - to live speech, and that a direct human voice improves their performance. This highlights how adaptable GWL dogs are in processing speech, but also suggests that a more natural context, in line with the spontaneous social interactions between humans and dogs may play a role in dogs' response to human speech.
Fumi Higaki, the study's lead author, expressed, "Initially, I was genuinely surprised to see dogs looking puzzled and struggling to respond to recorded words emitted on AIC buttons. As probably many people assumed, I believed that dogs perceive those the same way that I do. However, our results were robust and consistent across dog groups and tasks." "We hope our research will help improve methodologies in canine research that utilize recorded human speech," said project leader Dr. Claudia Fugazza. The authors also suggest that despite these interesting results, family dog owners should rely more on natural communication interactions rather than using arbitrary devices.
