New eye-tracking research sheds light on how dogs process and predict the actions of both human and non-human agents, demonstrating that they prioritize the location of an action over the identity of the object involved. This suggests a unique aspect of canine social cognition, where spatial information plays a crucial role in their understanding of others’ behaviors. The findings were published in the journal Animal Cognition.
Previous research has suggested that humans use their own motor experiences to interpret other people’s movements. This ability develops early in life and is seen in infants and apes. For instance, human infants and apes have been shown to anticipate that an agent will continue to interact with the same object, even if its location changes, suggesting a strong object-oriented approach in their action understanding.
However, dogs present an interesting case study due to their significant evolutionary and developmental differences from humans and apes. Despite their inability to replicate many human actions due to anatomical differences, dogs have a rich history of co-evolution with humans, making them keen observers of human behavior.
The researchers conducted the new study to test whether dogs, like human infants and apes, use their observations of actions to form expectations about future actions. Since dogs cannot physically perform actions like grasping or kicking, they do not have personal motor experience of these actions to rely on for making predictions. This makes them an excellent model for investigating whether understanding others’ actions requires the ability to perform those actions oneself.
“How do we understand what others are doing just by looking at their actions, without words needed? Psychologists have come up with different explanations. One hypothesis is that we use our own motor system to understand what our goal would be if we were executing the same actions that we now see someone else do,” explained study author Lucrezia Lonardo, a PhD candidate at the Messerli Research Institute of the University of Veterinary Medicine of Vienna.
“To test this hypothesis, it is extremely interesting to test a species (dogs) that does not have a motor system and motor representations similar to our owns, due to differing anatomy and locomotion, but that at the same time is routinely used to seeing our actions and successfully interact with us.”
The study was divided into two experiments, each designed to measure different aspects of the dogs’ anticipatory behavior and cognitive processing.
In both experiments, the researchers used video stimuli where two distinct objects were manipulated by either a human hand or an inanimate agent, such as a mechanical claw. The dogs were shown these videos on a screen while their head movements were restricted using a chin rest. This setup allowed for precise measurement of where the dogs were looking at any given moment.
The videos were structured in sequences of familiarization trials followed by test trials. During the familiarization phase, the dogs watched as the agent approached and interacted with one of the objects several times. After these repeated actions, the locations of the objects were switched. The test phase then measured the dogs’ anticipatory looks — where they expected the agent to go next.
The eye-tracking technology used in this study was capable of recording the dogs’ gaze at a high frequency, allowing researchers to determine exactly where the dogs were looking on the screen during the tests. Additionally, the researchers employed pupillometry — the measurement of pupil dilation — as a complementary measure to assess the dogs’ responses. Changes in pupil size were analyzed as indicators of the dogs’ engagement or surprise at the unfolding events.
One of the key findings was that dogs did not show an expectation that the agents would continue to approach the same object once its location had changed. Instead, their anticipatory gazes were more often directed toward the location where the action had previously occurred, regardless of which object was now present.
This suggests that dogs may be more sensitive to the location of an action rather than the specific object involved, highlighting a potential difference in how dogs interpret goal-directed actions compared to humans and other primates, who often focus more on the object of an action.
“It is possible that, when visually scanning actions, dogs, unlike what we humans do from approximately 6 months of age on, do not focus preferentially on the object people act upon but rather on the location where the action takes place. This is consistent with the idea that visual familiarity with an action is enough to successfully respond to it. Our finding is consistent with previous research suggesting that dogs pick up more easily words for actions (e.g., sit, come, etc.) rather than words indicating objects.”
Further insights were gained from the pupillometry data, which revealed that changes in the dogs’ pupil sizes corresponded more significantly to changes in the location of an action rather than changes to the identity of the object. Larger pupil dilations occurred when the agent approached a new location, suggesting that such events were more surprising or significant to the dogs.
The study also differentiated between the dogs’ responses to human and inanimate agents. Interestingly, while the anticipatory looking behavior did not significantly differ between these agent types, subtle differences were noted in the dwell times and first fixation data, suggesting that the type of agent could influence how dogs process and respond to actions. Dogs tended to have longer gaze durations and quicker fixations when observing human agents compared to inanimate ones, possibly due to the greater biological salience and familiarity of human movements.
These results contribute to our understanding of social cognition in dogs, showing that they can anticipate future actions based on previous observations, albeit with a focus on location over object identity. This ability to anticipate and react to the goal-directed actions of others, even when those actions do not directly involve them, is indicative of a sophisticated level of social awareness that supports effective communication and interaction with humans.
But the findings must be interpreted within the constraints of the experimental design, including the artificial setting and the limited sample size and diversity of actions tested. Future research could explore a broader range of actions and more varied contexts to better understand the flexibility and limits of dogs’ action anticipation abilities. Additionally, comparing these results with other species could further illuminate the evolutionary and developmental aspects of action understanding across different contexts and social environments.
“This was a study on spontaneous looking behavior, investigated using a somewhat ‘artificial’ setting in which dogs are trained (with positive reinforcement) to not move their head while watching videos on a computer screen. But of course it is possible to attract dogs’ attention to objects in an everyday interaction.”
“Eye-tracking is a non-invasive yet powerful method. Used within a comparative approach, it can offer a unique window into the minds of non-verbal participants: based on spontaneous eye-movements and pupil size measures it is possible to infer cognitive processes such as memory, expectations, and predictions. In particular, we would like to continue investigating dogs’ social cognition (how they perceive other agents, conspecifics or humans) and whether it resembles or diverges from human social cognition.”
The study, “Do dogs preferentially encode the identity of the target object or the location of others’ actions?” was authored by Lucrezia Lonardo, Christoph J. Völter, Robert Hepach, Claus Lamm, and Ludwig Huber.