Publications of Gergely, György

The attribution of navigational- and goal-directed agency in dogs (Canis Familiaris) and human toddlers (Homo Sapiens)

Both human infants and non-human primates can recognize unfamiliar entities as instrumental agents ascribing them goals and efficiency of goal-pursuit. This competence relies on movement cues indicating distal sensitivity to the environment and choice of efficient goal-approach. While dogs’ evolved sensitivity to social cues allow them to recognize humans as communicative agents, it remains unclear whether they have also evolved a basic concept of instrumental agency. We employed a preferential object-choice procedure to test whether adult pet dogs and human toddlers can identify unfamiliar entities as agents based on different types of movement cues specifying different levels of agency. In the Navigational Agency condition, dogs preferentially chose an object that varied it’s pathway to avoid collision with obstacles over another object showing no evidence of distal sensitivity (regularly bumping into obstacles). However, in the Goal-Efficiency condition where neither objects collided with obstacles as they navigated towards a distal target, but only one of them exhibited efficient goal-approach as well, toddlers, but not dogs, showed a preference toward the efficient goal-directed agent. These findings indicate that dogs possess a limited concept of environmentally sensitive ‘navigational agency’ that they attribute to self-propelled entities capable of modifying their movement to avoid colliding with obstacles. In contrast, dogs showed no evidence of attributing the higher-level concept of ‘goal-directed instrumental agency’ based on cues of efficient goal-pursuit. Toddlers, on the other hand, demonstrated clear sensitivity to cues of efficient variability of goal-approach as the basis for differentiating, attributing, and showing preference for goal-directed instrumental agency.

Seeing behind the surface: communicative demonstration boosts category disambiguation in 12-month-olds

In their first years, infants acquire an incredible amount of information regarding the objects present in their environment. While often it is not clear what specific information should be prioritized in encoding from the many characteristics of an object, different types of object representations facilitate different types of generalizations. We tested the hypotheses that 1-year-old infants distinctively represent familiar objects as exemplars of their kind, and that ostensive communication plays a role in determining kind membership for ambiguous objects. In the training phase of our experiment, infants were exposed to movies displaying an agent sorting objects from two categories (cups and plates) into two locations (left or right). Afterwards, different groups of infants saw either an ostensive or a non-ostensive demonstration performed by the agent, revealing that a new object that looked like a plate can be transformed into a cup. A third group of infants experienced no demonstration regarding the new object. During test, infants were presented with the ambiguous object in the plate format, and we measured generalization by coding anticipatory looks to the plate or the cup side. While infants looked equally often towards the two sides when the demonstration was non-ostensive, and more often to the plate side when there was no demonstration, they performed more anticipatory eye movements to the cup side when the demonstration was ostensive. Thus, ostensive demonstration likely highlighted the hidden dispositional properties of the target object as kind-relevant, guiding infants’ categorization of the foldable cup as a cup, despite it looking like a plate. These results suggest that infants likely encode familiar objects as exemplars of their kind and that ostensive communication can play a crucial role in disambiguating what kind an object belongs to, even when this requires disregarding salient surface features.