Skip to main content

Verified by Psychology Today


How Good Are Dogs at Recognizing Human Word Sounds?

Dogs, like humans, can recognize subtle distinctions in spoken word sounds.

Karl Witkowski (1860-1910) [Public domain]
Source: Karl Witkowski (1860-1910) [Public domain]

Dogs may have better linguistic talents than scientists have previously given them credit for.

The ability to recognize specific word sounds in human speech is a difficult task since small differences in sound frequencies can alter the meaning of a word. Consider the words "bat", "bit", "bet", and "beet". These words differ only in terms of their vowels, which results in a subtle change in the sound frequency between the consonants. These are the kinds of small sound differences that computer word recognition software often finds confusing.

Furthermore, speech sounds will change depending upon who is saying the words. This is because differences in body size, age, sex, mouth shape, and other factors change the pitch and other characteristics of the spoken sound. For example, it is remarkable how we can still recognize individual word sounds over a telephone since phone transmission eliminates much of the lower frequencies in the spoken word. Because of these complexities, a number of researchers have argued that spontaneous recognition of speech sounds is a complex and unique human ability, although with training it may be possible to get some animals, like dogs, to recognize certain specific words.

A team of scientists headed by Holly Root-Gutteridge at the School of Psychology at the University of Sussex decided to see just how good the spontaneous word recognition ability of dogs really is. First, they selected a set of words that all started with an "h" and ended with a "d" but varied only in terms of the vowel sound. Thus the set of words included things like "had", "head", "hid" and "hood". They specifically chose these words because they have no relation to typical commands that the dog might have learned to react to. The words were recorded from 13 male and 14 female speakers, none of whom were familiar to any of the dogs that would be tested. The speakers were not only of different sexes but also of different ages and had different accents.

The technique used in this research involves "habituation". This simply relies on the fact that when we sense something that is novel we tend to pay attention to it, but if it repeats a number of times it no longer catches our interest and we tend to ignore it. If something changes, however, it will break through our boredom and recapture our attention.

The basic experimental set-up involved a dog sitting with its owner near an audio speaker that would play a string of six of these recorded words, each separated by six seconds of silence. The dog's responses were videotaped and later scored.

One experimental trial might have run this way. The dog to be tested is presented with a string of repetitions of the word "had" through the speaker. Suppose that in this instance the word was spoken by a woman. Typically, when the dog first hears this new word spoken by this female voice he would point his ears forward, or move toward the speaker, or flick his eyes in the direction that the sound was coming from, all of which are signs of attention and engagement.

However, as other women with different accents repeat the word "had" the dog loses interest indicating that he knows that they are all saying the same thing. On the other hand, when a female speaker in the sequence says a new word, one with a different vowel, like "hid", the dog now perks up again, indicating that he noticed the difference. But when the next woman's voice returns to saying "had" his attention will again flag. Given this pattern of reaction in the tested group of canines, the experimenters concluded that dogs recognize spoken words irrespective of the speaker, and they don't need any training to do it.

These results are significant because they confirm two important aspects of speech recognition in dogs. The first, they can distinguish between subtle changes in vowel sounds that identify particular words. Second, dogs isolate the important word sounds from of all of the changes in sound quality associated with different speakers. This is a vital aspect of linguistic ability because without it you would not be able to understand what different speakers were saying simply because the words sound somewhat different when they are coming out of a different person, with a different accent, and a differently shaped voice apparatus.

Dr. Root-Gutteridge concluded, "This research shows that, despite previous assumptions, this spontaneous ability is not uniquely human and that dogs share this linguistic talent, suggesting that speech perception may not be as special to humans as we previously thought."

It is important to note, however, that given the nature of this set of tests in this study, the investigators can't demonstrate that the dogs actually "understood" what the words meant. However, it does show that the dogs are listening to us and trying to tease apart and recognize the speech sounds that we are making.

Copyright SC Psychological Enterprises Ltd. May not be reprinted or reposted without permission


Root-Gutteridge H, Ratcliffe VF, Korzeniowska AT, Reby D. (2019) Dogs perceive and spontaneously normalize formant-related speaker and vowel differences in human speech sounds. Biology Letters, 15: 20190555.

More from Stanley Coren PhD., DSc, FRSC
More from Psychology Today
More from Stanley Coren PhD., DSc, FRSC
More from Psychology Today