Dog Brains Pick Words Out of Speech the Way Human Brains Do, a First Outside Our Species
Hungarian scientists put EEG caps on 20 dogs and 20 people. Both locked on to words built from consonants, a bias never found even in monkeys, and even dogs raised on the street showed it.
Dogs break a stream of human speech into words using the same trick human brains rely on, according to a study published in the journal Science. It is the first time the pattern has been found in any species other than our own, including primates far more closely related to us.
The trick is called consonant bias. Words are built from vowels and consonants, and although vowels are louder, consonants form the skeleton of most words. From infancy, people tend to lean on consonants to pick individual words out of continuous speech. "In a continuous speech stream, it is easier to detect individual words when we focus on consonants," said Attila Andics, the cognitive neuroscientist who heads the Neuroethology of Communication Lab at Eötvös Loránd University in Budapest and led the study.
The question was whether that bias requires a uniquely human, language-ready brain, or whether it can grow out of simple pattern learning from hearing speech over and over. Dogs, which spend their lives surrounded by human talk but cannot speak, offered a test.
The team compared the brain responses of 20 people and 20 companion dogs using noninvasive EEG, which records electrical activity through electrodes on the scalp. Participants heard continuous streams of syllables with no pauses. In one type of stream, repeating consonant patterns formed three-syllable word skeletons while the vowels changed freely. In a second, repeating vowel patterns formed the words. A third, the control, had no repeating patterns at all. The researchers measured how strongly each brain fell into step with the rhythm of the words, a phenomenon called neural synchronization.
"In the consonant-structured speech streams, the dog brains, just like the human brains, successfully synchronized with the rhythm of the words," said Kinga G. Tóth, a doctoral researcher and co-first author. In both species, that synchronization was stronger than for vowel-based or random streams. About 400 milliseconds after each word began, both humans and dogs showed a distinct change in brain activity, but only for consonant-built words. That time window is known to reflect how brains segment speech into words.
The bias was not limited to pampered pets. Dogs that spent their first three months on the street or in a shelter showed an equally strong consonant preference. "This provides further evidence that consonant bias does not require prewired language-processing mechanisms that emerge during a specific early developmental period," Tóth said.
There were differences. People could also pick out vowel-built words, though less well, while dogs could not. And only human brains tracked the rhythm of individual speech sounds; for dogs, the syllable was the smallest unit they processed, suggesting people hear speech at finer resolution.
"Taken together, these findings suggest that the brain's tuning to speech, and specifically the emergence of consonant bias, does not require a prewired brain and humanlike language abilities," Andics said. "Instead, regular exposure to speech may be enough to change how the brain functions." The paper's co-first author is biologist Boglárka Morvai.
