Dogs’ brains track made-up words in a stream of speech more closely by their consonants than by their vowels, brain recordings show — a bias long documented in people — according to a study published Sept. 24, 2026, in the journal Science. The paper, from Attila Andics’s Neuroethology of Communication Lab at ELTE Eötvös Loránd University in Budapest, is titled “Neural evidence that dogs segment the speech they hear with a humanlike consonant bias.”
What the researchers tested
The researchers recorded brain activity with noninvasive scalp electrodes — EEG — in 20 adult humans and 20 companion dogs, according to reports on the study by Neuroscience News and Discover. Each listened to continuous streams of three-syllable nonsense words. In one stream, the invented words shared a repeating consonant pattern; in another, they shared a repeating vowel pattern; a third stream had no repeating pattern at all, according to Discover, which described it as “a stream with no pattern.” The researchers then measured two things, according to the paper’s abstract: how closely brain activity synchronized with the rhythm of the words (a measure called neural entrainment) and event-related potentials — brief spikes in electrical activity tied to a specific moment.
What they found
In both species, the brain synchronized more strongly with the words built on a consonant pattern than with those built on a vowel pattern, the researchers reported. About 400 milliseconds after each word began, both human and dog brains showed a distinct change in electrical activity — but only for the consonant-built words, according to Discover. The two species were not identical: human brains also tracked words built from vowel patterns, though less strongly, while dog brains did not, Discover reported. The paper’s abstract says the entrainment and event-related-potential measures “revealed enhanced word-level tracking for consonant- compared with vowel-structured streams in both species,” and concludes that “linguistic regularity–based processing biases can thus also emerge in a nonspeaking species exposed to speech.” In other words, an animal that does not talk can still pick up word patterns from the speech around it. Discover reported that co-lead author Kinga G. Tóth said, in a press release accompanying the paper, that this is the first evidence a consonant bias can emerge in the brain of a species other than humans. The bias also showed up in dogs with little early exposure to speech, Neuroscience News and Discover reported; Discover said this included dogs that appear to have spent their first three months on the street or in a shelter.
Why consonants matter
The paper’s abstract notes that across many human languages, consonants carry more lexical information than vowels — they do more of the work of telling one word apart from another. When people listen to a stream of speech with no pauses between words, as in ordinary conversation, the brain has to figure out where one word ends and the next begins. The abstract states that humans, unlike nonhuman primates, lean on consonant patterns more than vowel patterns to do this segmenting, “despite vowels’ greater acoustic saliency” — meaning vowels are the louder, more prominent sounds, yet the brain still weights consonants more heavily when parsing words. This study asked whether that same lopsided reliance on consonants, previously documented only in humans, also shows up in a species that is spoken to every day but does not produce words itself.
The lab’s earlier work
The Budapest group has been mapping how dogs’ brains handle human speech for a decade. In a 2016 Science paper, the same lab trained 13 dogs to lie still inside an fMRI scanner and found their brains processed word meaning and the intonation of speech in separate regions, with reward-related brain areas responding when praise words were spoken in a praising tone, according to an AAAS release at the time. In a 2022 study published in NeuroImage, 18 dogs that had heard only Spanish or only Hungarian from their owners showed different brain activity when they heard their familiar language compared with the unfamiliar one, according to a EurekAlert release from that January.
What it does not show
The new study measured brain responses to artificial streams of nonsense syllables, not to real words or sentences, and the paper describes this as evidence of segmentation — the brain finding where a word-like unit begins and ends — not comprehension. It does not show that dogs understand what people say to them, or attach meaning to the sounds they are tracking.