A dog identifies its owner through a closed door and from a recording. The recognition is more structured than a simple familiarity response, and the structure is now reasonably well described.
Two streams of information in one sound
A spoken phrase carries at least two separable things: the identity and emotional state of the speaker, and the word content itself.
These travel in different acoustic properties. Identity sits largely in the resonant qualities of the voice, while word content sits in the rapid changes across a syllable.
Dogs appear to process these separately rather than treating a phrase as one undifferentiated sound, which is why an unfamiliar person saying a known cue often produces hesitation.
Tone is handled before meaning
Work using imaging in trained, unrestrained dogs suggests intonation is processed in regions distinct from those responding to the words themselves.
Praise words spoken flatly, and neutral words spoken warmly, produce different patterns to praise words spoken warmly. Both channels contribute.
This is the mechanism behind the familiar observation that a dog can be insulted in a delighted tone and respond happily. Tone dominates when the two conflict.
Why high-pitched speech works on puppies
The exaggerated, high-pitched register people adopt with dogs is not merely sentimental. Young puppies orient to it markedly more than to ordinary adult speech.
The effect weakens with age. Adult dogs respond to it less reliably, and respond most to the voice of a person they already know regardless of register.
Familiarity gradually outweighs acoustic novelty, which is why a stranger's enthusiastic voice attracts a puppy but often does little for an older dog.
How recognition of individuals is built
Voice recognition in dogs seems to be learned by association rather than being present from birth. A voice acquires significance through what reliably follows it.
Dogs living in multi-person households distinguish between the residents, and their responses differ in ways that track what each person typically does with them.
Recognition also degrades in a predictable way. A voice heard rarely for a long period draws a weaker response than one heard daily, though it is rarely lost entirely.
Practical consequences for training
If identity and tone are processed alongside words, a cue trained by one person in one register is not automatically a cue at all when someone else says it.
Having every member of a household use the same word, and practising cues in a normal speaking voice rather than an excited one, removes a large source of apparent disobedience.
The same reasoning explains why a dog that responds perfectly at home ignores a stranger entirely. It is not defiance but a cue the dog has never actually been taught.