TDK conference 2019

The unique relationship between humans and dogs is supported by dog’s high efficiency in decoding human communication signals. Living in a human environment, dogs are constantly exposed to a continuous flow of speech. In similar conditions, infants quickly learn to compute the patterns present in the human speech, using a process called statistical learning. Dogs, although to a much lesser extent, are also capable of processing spoken language. While a few studies explored the word learning capacities of dogs, the neural basis of vocabulary acquisition, the cognitive mechanisms and the brain structures involved in this process are still unknown. Since statistical learning is a mechanism commonly found in the animal world, the aim of our research was to test whether it can aid dogs’ word learning, and to explore the corresponding brain areas.

Functional magnetic resonance imaging (fMRI) was used to examine if dogs could recognise the statistical distributions of sounds they hear in a continuous stream of speech. A voice generator was used to create 32 sec. long blocks, consisting of 1) syllables following each other in a randomly in the first condition and 2) certain syllables always coming together, thus forming artificial words in the second condition. 12 healthy, awake, unrestrained, fMRI-trained dogs were exposed to these, previously unknown stimuli. The experiment consisted of three parts, in all of which dogs listened to alternating blocks of the random and word condition: a baseline fMRI measurement, a learning phase and a second fMRI scan, when learning outcomes were tested.

In line with earlier dog brain imaging data, we demonstrated that the right and left mid ectosylvian gyri, as well as the right caudal sylvian and ectosylvian gyri are involved in the processing of acoustic stimuli. Importantly, our group results show that dogs can discriminate on the neural level between the random and the word condition. There was a significant interaction between repetition and condition in the rostral sylvian gyrus indicating, that the activity difference between the first and second random aquisition and the first and second word aquisition was significantly different (clusterwise pFWE < 0.05). Thus, in dogs, the ability for statistical learning of vocabulary items stems from the secondary auditory cortex, which is consistent with human data.

MRI and fMRI are becoming common in the veterinary diagnostics, particularly in patients where central nervous system involvement is suspected. By assigning specific anatomical structures to language processing, our results can catalyse diagnoses and highlight that certain changes in the dogs’ behaviour may result from alterations in the aforementioned auditory regions.

Based on our current results we are planning to examine brain areas involved in the recognition of the previously learned words using event-related fMRI in the next phase of the study.