But even if they can't reply, they may be listening. New research published in the journal Royal Society Open Science reveals why dogs appear to know we are speaking to them — yet only learn a handful of words over their lifetime. The key may lie in how dogs process changes in speech sound.
What they did — Researchers strapped a set of electrodes to 17 dogs's heads to monitor the electrical activity in their brains. Using this technique, also known as an electroencephalogram, or EEG, the researchers could track how the dogs' brains responded to different words. The brain response they were looking for in particular is a spike in activity, known as an event-related potential.
What's new — The researchers believe this may be the first example of using EEG to analyze word processing in dogs.
The technique can reveal whether dogs' brains process different sounding words in the same way or a different way, Lilla Magyari, study author and postdoctoral fellow at the Norwegian Reading Center for Reading Education and Research, tells Inverse.
Magyari and her colleagues tested the dogs on three types of words:
- KNOWN words dogs would recognize (e.g. "SIT")
- SIMILAR words which held no known meaning to the dogs ("SUT")
- NONSENSE words which were not similar at all to either of the words ("BEP")
To account for the possibility some dogs might be familiar with certain words more than other dogs in the study, researchers divided dogs into different groups based on how frequently owners used these words and dogs responded to them.
And then they compared each dogs' brain activity in response to different words.
What they found — "In our study, the dogs' [event-related potentials] were different when they listened to words they knew — e.g. 'sit' — and when they listened to very different nonsense words," Magyari says.
"This shows that dogs are capable of recognizing words they are familiar with — e.g. instruction words — just by listening to them," she explains.
This makes sense to human ears, too — the difference in sound between "Sit" and "Bep" is pretty significant, after all.
But what stumped the dogs most seemed to be when they listened to words which differed by a single speech sound — usually only one letter — changing the word's meaning.
"We also found that dogs' [event-related potentials] were not different when they listened to the words they knew, e.g.'sit,' compared to similar nonsense words," Magyari says. "This means that dogs can recognize a few words but they do not attend to details of how these words sound."
"This means that dogs could hear the difference between these words, but they probably do not pay attention to these smaller differences within words," she says.
What we don't know — These results were unexpected, Magyari says.
"Dogs have really good auditory capacity, and earlier experiments showed that they notice differences between different speech sounds — e.g. 'i' and 'a.' Therefore, it was surprising to find that dogs' brain do not differentiate words they know from nonsense words which differ only in one sound," she explains.
Why it matters — The EEG scans also revealed how quickly the dog's brain activity responded to each word — yielding some interesting dog-human comparisons.
Dogs' responses appear to track closely to those of an infant under the age of one — a critical period of language development, but one that generally goes on prior to the beginnings of verbal speech.
"Infants around 14 months do not process all details of words in certain experimental and word learning situations, similarly to dogs," Magari says. "Therefore, word processing of dogs is more similar to word processing of 14 month-old infants than to older infants or human adults."
Curiously, dogs' brains seem to respond at a similar pace to words which make no sense as humans' brains, distinguishing these "nonsense" words after only 200-300 milliseconds.
"This effect is in line with similar studies on humans which show that the human brain responds differently to meaningful and nonsense words already within a few hundred milliseconds," Magyari says.
Abstract: While dogs have remarkable abilities for social cognition and communication, the number of words they learn to recognize typically remains very low. The reason for this limited capacity is still unclear. We hypothesized that despite their human-like auditory abilities for analysing speech sounds, their word processing capacities might be less ready to access phonetic details. To test this, we developed procedures for non-invasive measurement of event-related potentials (ERPs) for language stimuli in awake dogs (N= 17). Dogs listened to familiar instruction words and phonetically similar and dissimilar nonsense words. We compared two different artefact cleaning procedures on the same data; they led to similar results. An early (200–300 ms; only after one of the cleaning procedures) anda late (650–800 ms; after both cleaning procedures) difference was present in the ERPs for known versus phonetically dissimilar nonsense words. There were no differences between the ERPs for known versus phonetically similar nonsense words. ERPs of dogs who heard the instructions more often also showed larger differences between instructions and dissimilar nonsense words. The study revealed not only dogs’ sensitivity to known words, but also their limited capacity to access phonetic details. Future work should confirm the reported ERP correlates of word processing abilities in dogs