It's one of the richest signals in the animal world — shaped by thousands of years living beside us, and structured enough that even software can read it. Here's what the research actually says about why dogs bark.
Adult wolves almost never bark — barking makes up only around 2.4% of their vocalisations, and what little there is comes out in short, isolated bursts. Dogs are the opposite: they bark readily, across dozens of situations, often in long, rhythmic stanzas.
Why the difference? Domestication. Dogs retain juvenile wolf traits into adulthood — a phenomenon called neoteny — and frequent barking is one of them. Living alongside humans selected for it: a dog that could call attention, raise an alarm, or simply "talk" to us had an advantage. In a real sense, dogs learned to bark for us.
Researchers describe barks with three acoustic ingredients — and these are exactly the features that carry meaning between dog and human:
In a striking finding, Pongrácz and colleagues showed that humans — even people who've never owned a dog — reliably read a dog's emotional state from the bark's acoustics alone. The mapping is consistent:
The 2am problem, explained. The barks people find most annoyingare low-pitched and rapidly repeated — and research suggests this combination is specifically tuned to grab human attention. The barks of a dog left alone are high-pitched and desperate; other dogs ignore them, but humans can't. If a neighbour's dog is keeping you up, that's not in your head — it's acoustics.
The landmark Budapest study recorded barks in six everyday situations and found each had a recognisable acoustic signature — different enough that a machine-learning model could tell them apart:
Someone approaching — alert, watchful.
Conflict or threat — low and harsh.
Anticipating a walk — excited.
Left by themselves — high, repetitive.
Wanting a toy — insistent.
Play invitation — bright, tonal.
In 2008, a machine-learning model classified those 6,000+ barks by context— and even by individual dog — outperforming human listeners on some tasks. A 2024 deep-learning study pushed it further, training on 19,643 barks from 113 dogs.
barker stands on exactly this work. The free tool runs YAMNet — a Google sound-recognition model — entirely on your device, to tell a real dog bark from the TV, a door, or a person talking. The audio never leaves your browser; only the pattern of barks is kept.
Privacy by design. Because detection happens in your browser, barker works from the derived pattern — bark times, lengths, and loudness — not a recording. Human speech is filtered out before anything is saved.
Veterinary behaviourists group most barking into a handful of causes. Knowing which one you're hearing is the first step to helping:
Record about a minute and our on-device model gives you a playful, honest read on your dog's barking — free, no signup.