Object connectedness affects quantity perception in dogs.

Research in humans has shown that connecting a collection of dots with task-irrelevant lines leads to a robust numerical underestimation. To date, this paradigm has not been explored in other animal species. We assessed if connectedness results in an underestimation of quantity in dogs. Dogs (N=30) underwent an initial task, where they were presented with two plates containing different numbers of food items, and were trained to choose the stimulus containing the larger amount of food. Once they reached the learning criterion, dogs were presented with two arrays, each of which contained 30 food pieces, arranged in identical locations. On one array the food pieces were connected by thin inedible lines, forming three clusters of ten elements each. On the other array, identical lines were arranged randomly amongst food elements without being in contact with them. Each dog was presented with the test condition four times, and were expected to maximize food intake, choosing the stimulus where food items appeared more numerous. Thirteen dogs completed the test, the rest of the dogs did not reach the learning criterion in the training phase. An intercept-only GLMM model on dogs’ choice in test trials (connected or unconnected stimulus) revealed that dogs chose the unconnected stimulus more often than expected by chance (35/52, p=0.015). This suggests that connectedness affects quantity judgements by dogs in a similar way to humans. The result is of particular relevance in view of the fact that previous research generally failed to find effects of non-numerical cues on quantity estimation by dogs. Furthermore, dogs are the first non-human species in which the illusion has been assessed. In the light of the scarce susceptibility to quantity illusions by many animal species, it would be interesting to know if susceptibility to this illusion applies also to other species