A team of US scientists on the Dog Project, have been studying how dogs process the words that are spoken to them by their owners. The scientists have trained the dogs they study to enter into an MRI scanner voluntarily, without sedation or restraint and stay still.. They have already examined subjects such as the dogs’ neural responses to expected reward, identified the areas in a dog’s brain that help to process faces, shown the dogs’ olfactory responses to both humans and dog odours and helped to link prefrontal brain function to inhibitory control. This study was published in Frontiers in Neuroscience, a peer-reviewed magazine. The article is available to read online.
The scientists stated that although dogs are able to follow verbal commands, they may also rely on other cues, including gestures, their owner’s gaze direction or emotional expressions to help them follow that command. They wanted to examine which brain mechanisms were used by the animals to show what constitutes a word to a dog or what enables them to differentiate between words.
Twelve varied breeds of dog were recruited for the study and trained to retrieve two different objects, based on their names, by the owners. The training took months for the dogs to be secure in their understanding of the words. The objects were of different textures to help the animals differentiate between the two, for example a rubber and a soft toy. The dogs were asked to fetch one of the objects and then rewarded with food or praise. The dog had to prove that it could consistently choose the right toy when asked and shown both objects. Once the dog could choose one of the objects 80% of the time, and the other object over 50% of the time, it was asked to perform the tasks for the study. Training the dogs took between two to six months, using bi-weekly practices. The dogs had all previously taken part in other studies using the fMRI scanner.
The dogs were placed in the scanner and the dog’s owner spoke the words. Both owners and dogs wore ear plugs to help reduce the scanner noise, but not enough to block out the sound of words. The scientists controlled which words were said and when they were presented to the owner. The dogs could leave the scanner if they wished. Four types of trial were offered to the dogs, expected words, unexpected words, reward and pseudo words, which were generated randomly, but could be changed if they were too close to the object word. For expected words, the dogs were offered the expected object and allowed to interact with it. Unexpected enabled the owner to speak an expected word, but offer a different object. Pseudo word trials paired a pseudo word with a different object and reward trials were interspersed where the owner rewarded the dog for staying in the scanner with food. The dogs all took part in the same trial sequence.
The dogs were able to select object1 on 80-100% of the trials, apart from one and selected object2 correctly on 60-100% of the trials. The results were significantly better than chance and the dogs were able to perform better on word1 than word2. The scientists found that the dogs could process human speech by differentiating between words they had heard before and those they had not, but they showed the greatest response to novel words. The researchers suggested that the pseudo words had not been presented so often, or the dogs had not been able to attach meaning to them.
Humans respond with brain activation to words that they know better than pseudo words. More brain processing can be seen when pseudo words are similar to a known word. The dogs may also reacted to the similarity between words they knew and those they did not, but it has been previously shown that dogs are able to tell the difference between altered phonemes of well-known commands. Humans learn nouns first, while dogs are mostly exposed to verbs.Humans are able to generalise the meaning of words in a number of contexts, but dogs tend to understand the word in the context in which it is trained. The scientists suggested that the dogs’ neural network reacted more strongly to unknown words, because it felt its owner wanted it to learn the word, so it paid close attention to it.
The dogs also showed activation in different parts of the brain during the tasks. The scientists suggested that this could be because the dogs were different breeds and sizes, which can have variations in intelligence. The size of the study was also a limitation.
Although humans tend to use verbal cues with their dogs, because that is how they, themselves, best learn, this study suggests that using visual and scent cues could also help the dog to learn faster.
Berns, G.S., et al., Awake fMRI Reveals Brain Regions for Novel Word Detection in Dogs, Frontiers in Neuroscience, October 2018